test(wiring): compile the 15 test files Cargo never built

[swackhamer]

The defect

Cargo compiles only tests/*.rs as test roots. Files in subdirectories are reached solely through mod / #[path] declarations. 21 files under tests/ were declared by nothing, so 114 #[test] functions had never once compiled — let alone run.

tests/integration/mod.rs was itself unreachable: nothing declares mod integration;. Every module it listed was therefore dead too.

An undeclared test file is indistinguishable from a comment. This is how the fabricated registries/apple.rs survived — apple_makernotes_tests.rs asserted its invented tag names and never built.

This PR

Revives 15 of the 21 files (24 tests). The 5 vendor MakerNote files follow separately; tests/ffi/build.rs is a build script, not a test module.

file	tests	result
error_handling_tests.rs	9	10 pass (one added)
{mkv,webm,flv,avi,mts}_integration_tests.rs	5	pass
{mp3,flac,aac,wav,ogg,opus,ape}_integration_tests.rs	7	pass
unit/audio/mp3_tests.rs	2	pass

tests/integration/mod.rs is removed: it duplicated two entries already in tests/integration.rs, and its 12 container modules are now declared there like every other integration module.

Three defects the dead code was hiding

1. API drift. All 12 container tests called MetadataMap::from_file (no longer exists) and stringified via TagValue::to_string() (never existed — TagValue has no Display). Rewritten to the current BufferedReader + parse_<fmt>_metadata entrypoints and the explicit variant match already used by infiray_tests.rs.

2. The parity comparison compared nothing. The tests ran exiftool -json but looked up group-qualified keys like FLAC:SampleRate. Plain -json emits bare tag names, so every lookup missed. Eleven files hit a continue and passed vacuously; flac_integration_tests.rs lacks that guard and compared "44100" against the literal string "null". Fixed with -G0, which emits exactly the asserted keys. FLAC now genuinely compares SampleRate/Channels/BitsPerSample against ExifTool, and matches.

3. A test with a wrong premise. test_error_unsupported_format fed in "INVALID_FORMAT_MAGIC_BYTES_HERE" and asserted it was not a known format. But that is printable ASCII, and ExifTool 13.59 reports it as FileType: TXT, MIMEType: text/plain — oxidex answering TXT was correct. The test now uses genuinely unrecognized binary (ExifTool 13.59: Error: Unknown file type), and a new sibling test test_printable_ascii_is_txt_not_unknown pins the TXT behaviour so a future change to it reads as the regression it would be.

Separately, test_error_truncated_tiff required the error text to contain one of eof/truncate/unexpected/invalid. The parser correctly reports IFD offset 8 exceeds file size 8 — a precise truncation diagnosis matching no substring. Widened the accepted set rather than changing a correct parser.

Verification

• No test deleted.
• cargo test --all-features --test integration --test unit_tests → 554 passed, 17 passed, 0 failed.
• cargo test --test integration -- --ignored *_integration_tests → 12 passed, 0 failed.
• Ground truth: ExifTool 13.59 via /usr/bin/perl5.34 -I/tmp/oxidex-exiftool-cache/exiftool/lib.
• cargo fmt --all and cargo clippy clean.

🤖 Generated with Claude Code