APP14 Marker, Better Subsampling and High Quality Improvements

[jonnyawsom3]

Description

The APP14 tag is now added to RGB(XYB), CMYK and YCCK JPEGs, as it is required in some decoders.
Using --chroma_subsampling will now correctly subsample XYB's Blue channel.
XYB is now 444 by default to allow JPEG XL Transcoding.
YCbCr now uses 444 by default and 420 at Quality 80/Distance 1.9 or lower.
Adaptive Quantization is disabled at Quality 90/Distance 1 or lower for non-XYB.
RGB is now used at Quality 100/Distance 0 for RGB input (PNG), overridden by XYB.

Fixes #122
Fixes libjxl/libjxl#3512
Fixes libjxl/libjxl#3605

Pull Request Checklist

• CLA Signed: Have you signed the Contributor License Agreement (individual or corporate, as appropriate)? Only contributions from signed contributors can be accepted.
• Authors: Have you considered adding your name to the AUTHORS file?
• Code Style: Have you ensured your code adheres to the project's coding style guidelines? You can use ./ci.sh lint for automatic code formatting.

Please review the full contributing guidelines for more details.

[monad0]

consider removing personal notes and fixing comment formatting

[Galaxy4594]

Build issues should be solved, I've managed to successfully compile on my Ubuntu VM.

[jonnyawsom3]

Marking as Draft while we do testing and cleaning up.

[ampaze]

I am eagerly awaiting the fix for the APP14 Marker (to make XYB work with Safari). Is there something missing or could this be merged?

[ampaze]

Testing this PR I noticed that just adding --xyb creates files that are bigger than without. I was under the impression that using XYB should generate smaller files?

I am using the following commands

cjpegli input.png output.jpg -d 4.5 --chroma_subsampling 420
cjpegli input.png output-xyb.jpg -d 4.5 --chroma_subsampling 420 --xyb

[jonnyawsom3]

Split into separate PRs #135, #136 and #137.

[jonnyawsom3]

Testing this PR I noticed that just adding --xyb creates files that are bigger than without. I was under the impression that using XYB should generate smaller files?

I am using the following commands

cjpegli input.png output.jpg -d 4.5 --chroma_subsampling 420
cjpegli input.png output-xyb.jpg -d 4.5 --chroma_subsampling 420 --xyb

XYB only subsamples the Blue channel, and is most likely higher quality than YCbCr. The difference between them could be due to a number of factors, including the image content itself.

[ampaze]

XYB only subsamples the Blue channel, and is most likely higher quality than YCbCr. The difference between them could be due to a number of factors, including the image content itself.

I was suspecting something like that. It would have been more intuitive if the distance parameter was consistent between the two modes.