fix(airplay): restore the transient pairing and control channel work lost in #42 - #43

What happened

#41  11:40:23Z  master ← fix/airplay-hkp-header
#42  11:40:35Z  fix/airplay-hkp-header ← fix/airplay-srp-proof-hash-g

#42 was stacked on #41. Twelve seconds after #41 merged that branch into master, #42 merged into the same branch — which master had already taken and moved past. Both merges reported success and both PRs show as merged, but #42's commits landed on a branch nothing points at.

Why it matters

Master currently has the X-Apple-HKP header from #41 and none of what it was a prerequisite for:

• the SRP proof still hashes g padded → pair-setup fails at M4 with HAP error 2
• transient pairing still runs M5/M6 → the receiver closes the connection after a successful M4
• there is no encrypted control channel at all

So AirPlay pairing on master is broken in exactly the way #42 fixed, while looking like it was fixed. This restores it.

Why this is a replay, not a merge

The obvious repair — merge fix/airplay-hkp-header into master — is wrong and destructive. That branch was cut before #35, #36, #37, #39 and #40 merged, so its diff against master reads as:

crates/openplay-discovery/src/address.rs      | 135 -------   ← deletes #37 entirely
crates/openplay-sender/src/app.rs             | 244 ++-----   ← reverts #35 and #40
crates/openplay-pipeline/.../airplay_...rs    |  21 +-        ← reverts #36
crates/openplay-receiver/src/main.rs          |  20 +-        ← reverts #39

Merging it would silently undo five landed fixes to repair one. Only the range between #42 and its own parent is safe to replay, and that range touches nothing outside openplay-airplay:

 examples/control_probe.rs |  79 ++++++
 examples/pair_probe.rs    |  29 ++-
 src/control_channel.rs    | 213 ++++++++++++++
 src/hap_pairing.rs        |  55 +++-
 src/http_session.rs       |  46 ++++
 src/lib.rs                |   1 +
 src/session.rs            |  59 ++---
 src/srp.rs                |  16 +-

That is what this PR contains — #42's content, unchanged, on top of current master.

Verification

• cargo test --all — 233 passed / 0 failed
• clippy --all-targets --all-features -D warnings clean, fmt --check clean
• Explicitly confirmed still present after the replay: openplay-discovery/src/address.rs (fix(discovery): sort discovered addresses by connectability #37), the row contrast fix (fix(sender): make receiver names readable on the unselected row card #35), the Miracast port field (feat(sender): let a manually added Miracast receiver specify its port #40), and the h264parse reordering (fix(pipeline): link the byte-stream capsfilter after h264parse, not before #36)

Note for future stacked PRs

Merging a stack bottom-up in quick succession does this silently. Either merge the top of the stack first, or re-target the child PR to master and let GitHub recompute its diff before merging.