ARW: add Sony "ARW 6.0" lossy decoder (compression 32766; A7R VI, A7 V)

[Kab1r]

Summary

Adds SonyArw6Decompressor, implementing Sony's "ARW 6.0" lossy codec
(TIFF Compression 32766) used by the A7R VI (ILCE-7RM6) and A7 V
(ILCE-7M5) in their Compressed / Compressed HQ RAW modes. These formats were
previously undecodable.

What this adds

• SonyArw6Decompressor (+ SonyArw6LogToLinear LUT) — the codec itself.
• compression == 32766 dispatch in ArwDecoder, the active-area crop, and a
  dimension-ceiling bump for the 66.8 MP sensor (which also enables the A7R VI
  lossless tiled LJpeg path).
• ILCE-7RM6 camera entry in cameras.xml. (The ILCE-7M5 entry already
  exists.)
• A fuzz target for the new decompressor.

Pipeline

Container tiles → per-row GCLI bitplane entropy decode → inverse reversible
5/3 wavelet (LL0 + 3 detail levels) → log→linear LUT → per-line colour
conversion → YCC→RGGB mosaic compose.

Coefficients are int16; tiles and their (up to four) components decode in
parallel under OpenMP (a no-op when it's unavailable). The per-tile sub-header
marker carries a 1-byte format tag that varies by body ('0' on the A7R VI,
'A' on the A7 V), so only its fixed "000" suffix is matched; the
black/white-level domain (1024/32800) and the DefaultCrop-derived active area
are codec-universal.

Reverse-engineering & validation

The format was reverse-engineered clean-room by observing the decoder's
input/output behavior; the 4096-entry log→linear table was recovered as a
black-box input→output sampling, not transcribed.

The full pipeline is validated byte-exact (MAE 0) against the reference
decoder output (Adobe DNG) on both bodies, full-frame and APS-C, in both lossy
and HQ modes.
Malformed input is rejected via ThrowRDE rather than crashing,
hanging, or invoking undefined behavior (a fuzz target is included).

I have also personally tested the decoder with a dirty build of Darktable built against this branch and was able to edit images successfully.

AI usage

This contribution was produced with
substantial AI assistance. The AI was used as a tool by me; the codec was
reverse-engineered from, and validated stage-by-stage against, the reference
decoder output, and I take full responsibility for the change and stand behind
its correctness.

[github-actions]

The proposed diff is not clang-formatted.
To make this check pass, download the following patch
(via browser, you must be logged-in in order for this URL to work),
(NOTE: save it into the repo checkout dir for the snippet to work)
https://github.com/darktable-org/rawspeed/actions/runs/28281503790/artifacts/7922007367
... and run:

cd <path/to/repo/checkout> # NOTE: use your own path here
unzip clang-format.patch.zip
git stash # Temporairly stash away any preexisting diff
git apply clang-format.patch # Apply the diff
git add -u # Stage changed files
git commit -m "Applying clang-format" # Commit the patch
git push
git stash pop # Unstast preexisting diff
rm clang-format.patch.zip clang-format.patch

[github-actions]

The proposed diff is not clang-formatted.
To make this check pass, download the following patch
(via browser, you must be logged-in in order for this URL to work),
(NOTE: save it into the repo checkout dir for the snippet to work)
https://github.com/darktable-org/rawspeed/actions/runs/28297440066/artifacts/7926886445
... and run:

cd <path/to/repo/checkout> # NOTE: use your own path here
unzip clang-format.patch.zip
git stash # Temporairly stash away any preexisting diff
git apply clang-format.patch # Apply the diff
git add -u # Stage changed files
git commit -m "Applying clang-format" # Commit the patch
git push
git stash pop # Unstast preexisting diff
rm clang-format.patch.zip clang-format.patch

[LebedevRI]

I should probably post an issue documenting the current maintenance stance of this project.
(but the TLDR is: i'm not interested in touching C++ any more. at least for free.)