…ield references

TypedEventTests asserted that argument-validation error messages
contained the C# property name being checked: Does.Contain("Status"),
Does.Contain("Flow"), Does.Contain("Currency") (twice),
Does.Contain("Amount"), Does.Contain("Value").

Each inline string mirrored a Progression / Resource / Purchase
property name. A property rename in the SDK (Resource.Currency to
Resource.CurrencyCode, say) would break the message wording but
pass the test, since "Currency" was matched as a substring even
when the SDK no longer mentioned it.

Replaces six inline strings with nameof(Progression.Status),
nameof(Resource.Flow), nameof(Resource.Currency), nameof(Resource.Amount),
nameof(Purchase.Currency), nameof(Purchase.Value). Each now tracks
the property name at compile time; renaming the property either
updates the test automatically or fails to compile.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

LogTests had inline literals for the Log.Debug / Log.Warn inputs
("silent", "hello", "something off"), the prefix substring asserted
on emitted lines ("[ImmutableAudience]"), and the warn marker
("WARN").

Adds four file-local consts (SilentDebugInput, EnabledDebugInput,
WarnInput, WarnMarker) at the top of the fixture and migrates the
six call sites. The prefix assertion now references Log.Prefix, the
SDK's existing const for the same string, so a SDK-side prefix
rename automatically propagates.

The "WARN" substring is kept as a local marker rather than slicing
Log.WarnPrefix; "[ImmutableAudience] WARN:" includes punctuation
the assertion does not require. A direct WARN constant matches the
test's actual intent ("the warn pipeline includes WARN somewhere").

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

ImmutableAudienceTests was using inline strings to assert against captured log lines and queue file
contents:

- Has.Some.Contains("Dropping") for the dropped-event marker that Track / Identify / Alias warns
  share
- "\"purchase\"", "\"identify\"", "\"alias\"" for queue-file content checks (the wire-format
  envelope's "type" / "eventName" fields)

Adds AudienceLogs.DroppingMarker = "Dropping" so the marker has a single source of truth and the
test references it directly.

Migrates queue-file content checks to interpolate against the SDK's existing EventNames.Purchase,
MessageTypes.Identify, and MessageTypes.Alias constants so a backend rename to those wire strings
would touch one place rather than scatter across the test.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

…ic user ID fixtures

ImmutableAudienceTests had four inline ID fixtures around the
identity / alias write tests:

- "76561198012345": real-shape Steam community 64-bit ID, used as
  the Identify input AND as the queue-content assertion subject
- "user1": minimal generic userId for the discarded-identify test
- "steam123": generic Steam ID for the alias-write test, used as
  both Alias input and queue-content assertion subject
- "user_456": generic Passport ID for the alias-write test

Adds SteamId64, GenericUserSingleId, SteamId, PassportId to
TestFixtures and migrates the six call sites. The SteamId64 / SteamId
input-assertion pairs were the worst offenders since a typo on one
side would silently pass on the other.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

…StoreTests inline JSON keys

SessionTests had "track explode" inline four times as the
InvalidOperationException message thrown by ThrowingTrack delegates
across four session lifecycle scenarios. Adds a TrackExplodeMessage
file-local const at the top of the fixture and migrates all four.

DiskStoreTests built a Purchase-shaped JSON payload by hand-rolling
the wire format with inline keys ("type", "eventName", "anonymousId",
"userId", "properties", "currency", "value") and inline values
("track", "purchase", "USD", "a"). Replaces those with the SDK's
existing constants (MessageFields.*, MessageTypes.Track,
EventNames.Purchase, EventPropertyKeys.Currency / .Value) and the
TestFixtures fixtures already centralised (UsdCurrency,
TestEventNames.PlaceholderA).

The "u" userId placeholder is left inline; centralising a single
one-char placeholder would add more noise than it removes.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

Renames TestFixtures.DistributionPlatformSteamCased and DistributionPlatformSteamUppercase to
neutral SteamPascalCase and SteamUpperCase. The same "Steam" / "STEAM" string fixtures are relevant
to two test families: ImmutableAudienceTests' DistributionPlatform lowercase normalisation, and
IdentityTypeTests' ParseLowercaseString case-insensitive matching. The new names drop the
family-specific prefix.

Adds:
- PassportPascalCase ("Passport") for the IdentityType mixed-case fixture batch.
- SteamSuffixed ("steamX") for the steam-prefix-but-not-exact-match fixture in the Custom-fallback
  test.

Migrates references:
- ImmutableAudienceTests:1127 / 1138 (rename only)
- IdentityTypeTests:55-57 (ParseLowercaseString_AcceptsMixedCase cases)
- IdentityTypeTests:66 ([TestCase("steamX")])
- ImmutableAudienceTests:398 (missed inline "user1", now GenericUserSingleId from the earlier
  centralisation pass)

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

…ort, stress, and offline tests

Three test files had inline publishable-key fixtures with no
intra-test reason for diverging from the canonical
TestDefaults.PublishableKey:

- ThreadSafetyStressTests:48 used "pk_imapik-test-stress"
- OfflineResilienceTests:47 used "pk_imapik-test-key"

Both fall under "any test-prefix key, content does not matter".
Migrated to TestDefaults.PublishableKey.

HttpTransportTests:162 used "pk_imapik-prodkey" specifically to
verify a non-test-prefix key resolves to ProductionBaseUrl. The
test cannot use TestDefaults.PublishableKey since that one IS
test-prefixed. Added a file-local ProdPublishableKey const
alongside the existing response-body fixtures and migrated the
one usage. The const carries a comment explaining the prefix
constraint.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

HttpTransportTests had "Retry-After" inline three times (across the
delta-seconds, http-date, and past-date 429 tests) and ConsentSyncTests
had it once in CapturingHandler.SendAsync. The string is the
RFC 7231 standard header name; the SDK reads it via the typed
response.Headers.RetryAfter accessor, but the mocks set it by name.

Adds a file-local RetryAfterHeader const at the top of each fixture
(matching the existing publishable-key / response-body file-local
const pattern) and migrates all four sites.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

…d values

JsonTests had inline scenario fixtures across nineteen tests:
- "flag" (Bool true / false serialise pair)
- "n" (Int / Long serialise)
- "x" (Null serialise)
- "v" (eight float / double NaN, Infinity, normal-range, large /
  small exponent tests; reused as the diamond test's inner value)
- "items" (List serialise)
- "level", "score", "perfect", "tags" (RealisticEventPayload nested
  properties keys)
- "fast", "clean" (tags array element values)
- "next" (deep-nesting MaxDepth guard)
- "self" (cycle-detection self-reference)
- "cycle" / "nesting exceeds" (FormatException message markers)
- "k", "a", "b" (diamond shared-child scenario)

Adds a file-local const block at the top of the fixture grouping the
literals by scenario (Bool / Numeric / Null / V / Array,
RealisticPayload, cycle / depth, diamond) and migrates every test
that consumed them. Where an SDK / fixtures const already covered
the same string (TestEventNames.PlaceholderA for "a",
TestEventNames.LevelComplete for "level_complete"), the test now
references that instead.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

…malformed inputs

JsonReaderTests had inline scenario fixtures across the deserialise
tests:
- "small" / "big" (IntAndLong)
- "t" / "f" / "n" (BoolAndNull)
- "arr" / "two" (Array)
- "abc" (RoundTripViaSerializer anonymousId)
- "76561198012345" (RoundTripViaSerializer userId; same string also
  used in ImmutableAudienceTests' Steam Identify test)
- "{not valid}" / "{\"a\":}" / "{\"a\":\"unterminated" (three
  malformed inputs in MalformedThrows)

Adds a file-local const block at the top of the fixture grouping the
keys, the array-element value, the anonymousId placeholder, and the
three malformed inputs. Each test now interpolates the encoded JSON
form from its key consts so a key rename touches one place. The
"76561198012345" duplication is removed by referencing
TestFixtures.SteamId64 (already centralised).

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).

EventQueueTests' local Msg helper labelled its dict's event-name
field with an inline "event" key. The string is distinct from the
SDK's wire-format MessageFields.EventName ("eventName"); EventQueue
accepts arbitrary dicts so the test could pick anything. Inline made
this implicit.

Adds a file-local MsgEventKey const at the top of the fixture with a
comment noting the test-scaffolding role and migrates the one usage.

Per the user's "everything random goes in a constant" stance.

Follow-up to SDK-272 (centralisation of duplicated literals).