fix(decoders): report completion from terminal states instead of stalling

[MagicalTux]

Follow-up to #123, which fixed this class of bug in deflate/deflate64/zlib. Reviewing all 53 raw_decode implementations found the same defect in seven more codecs.

The defect

A decoder parked in its terminal state that returns done: false with nothing consumed and nothing written — while the caller still holds input — is mapped by the RawDecoder→Decoder bridge to Status::OutputFull, i.e. "drain and call me again", with nothing left to progress on. Any caller looping until StreamEnd spins: CPU-bound, no allocation, so neither an output-size cap nor a memory cap can stop it.

Fixed

codec	why it stalled
zstd, xz	Done arms returned false; both terminal (neither supports concatenated frames/streams)
lz4 block, lzo	Done after the zero-length terminator block
lz4 frame	only reachable by calling past StreamEnd, but the state is terminal either way
lzx, amiga_lzx	Done arms returned false
rar5	returns on State::Done before the block that accepts caller bytes — a container passing the next header after the payload could never make progress
gzip	see below

gzip is the subtle one. Its BetweenMembers arm sets phase = Done intending to reach the Done arm that swallows trailing bytes (added precisely to avoid this spin). But match arms do not fall through, so the loop's no-progress check returned first and the swallow never ran. It needs an explicit continue, like the adjacent 0x1F path already has.

Two call shapes

Only the first was covered before:

1. trailing bytes present in the same call
2. trailing bytes arriving after the payload completes — what a container does when handing over the next header

xz and gzip pass shape 1 and failed shape 2; both looked safe until the second shape was tested. Tests cover both.

What is not a bug

Codecs that only ever report InputEmpty and produce output on finish (brotli, lzma, and the other buffering decoders) are correct as-is — the trait permits it and it terminates. The invariant asserted is therefore not "must report StreamEnd" but "must never report OutputFull with zero progress".

Verification

Each fix was found by reading the state machine and confirmed with a probe driving real encoder output plus trailing bytes. A sweep over all 50 round-trippable algorithms now reports no stalls. tests/terminal_state.rs covers every fixed codec plus controls; rar5 is decoder-only so it gets equivalent tests against its own fixture.

cargo test --all-features: 1766 pass, 0 fail. fmt and clippy -D warnings clean.