fix(makernotes): correct Apple MakerNote PrintConv tables to match ExifTool 13.59

[swackhamer]

Summary

Five Apple MakerNote PrintConv tables in src/parsers/tiff/makernotes/apple.rs disagreed with ExifTool 13.59's Apple.pm, found while reviving unreachable test files (PR #404 chain). Ground truth: Apple.pm from the ExifTool 13.59 source tree.

• DECODE_IMAGE_CAPTURE_TYPE (0x0014 ImageCaptureType) — worst offender, disagreed on every shared key. 10 (common on modern iPhones) printed "Screenshot" instead of "Photo"; 1 printed "Portrait" instead of "ProRAW"; 2 printed "Panorama" instead of "Portrait". ExifTool's 11/12 were missing. Trimmed to ExifTool's {1,2,10,11,12}.
• DECODE_CAMERA_TYPE (0x002e CameraType) — omitted 0 "Back Wide Angle" and invented five values. Trimmed to ExifTool's {0,1,6}.
• DECODE_OIS_MODE (0x000f OISMode) — Apple.pm declares no PrintConv at all (# seen: 2,3,5); ExifTool prints the raw integer. Removed the invented decoder; the tag now prints the raw value.
• DECODE_GREEN_GHOST_MITIGATION (0x003F) and DECODE_SNR_TYPE (0x0026) — both Unknown => 1 in Apple.pm with no PrintConv, so ExifTool omits them by default (only shown with -u). Both tags are now suppressed, mirroring the existing Unknown => 1 handling in sigma.rs.

Also caught by auditing the rest of the file's const_decoder! tables:

• DECODE_SEMANTIC_STYLE (0x0040) had no basis in Apple.pm either — the real tag uses ValueConv => \&ConvertPLIST (binary plist), not an integer enum — so its five invented labels are replaced with the raw extracted value.
• DECODE_PORTRAIT_MODE, DECODE_SCENE_TYPE, DECODE_LENS_MODEL were dead code (defined, never called from parse_entry, zero correspondence to any Apple.pm tag) — deleted outright.

This folds in the already-fixed DECODE_HDR_TYPE (0x000a HDRImageType) from fix/salvage-apple-samsung-makernote-tests, since both trace to the same root-cause audit.

Per AGENTS.md: a plausible-but-wrong value under a real ExifTool tag name is worse than an absent tag, since nothing downstream can tell it's wrong — omit rather than approximate. Unmapped PrintConv values now fall through to "Unknown (N)" instead of an invented label, matching ExifTool's own unmapped-value behavior.

src/parsers/tiff/makernotes/registries/apple.rs's doc comments and #[cfg(test)] assertions (which pinned the old fabricated values) are corrected alongside the tables, and the raw-fallback/suppression tags are switched from register_simple_i16 to register_raw.

Test plan

• cargo build --lib — compiles clean
• cargo test --lib apple:: — 26/26 passing (includes new tests pinning unmapped values to "Unknown (N)", raw-value printing for OISMode/SemanticStyle, and suppression for GreenGhostMitigationStatus/SignalToNoiseRatioType)
• cargo fmt --all
• cargo clippy --all-features -- -D warnings — clean

[swackhamer]

Closing without merging: this branch is now fully superseded by #394 ("find the MakerNote IFD, then read its binary plists"), which landed on `main` after this PR was opened.

#394 is a complete rewrite of `apple.rs` that already fixes every issue this PR addressed — HDRImageType (3/4), ImageCaptureType (1/2/10/11/12), CameraType (0/1/6), OISMode (no PrintConv, raw value), and correctly omits the `Unknown => 1` tags (`SignalToNoiseRatioType`, `GreenGhostMitigationStatus`, etc.) — and goes further: it implements real binary-plist decoding for `SemanticStyle`/`SemanticStyleRenderingVer`/`SemanticStylePreset` (0x0040/41/42) via a new `binary_plist` module, which was still an open gap in this PR. It also fixes an unrelated IFD-offset bug (14-byte header) that this PR never touched.

Rebasing this PR onto current `main` would just be re-deriving #394's work through a worse, table-less architecture. Closing in favor of it.