Note the sRGB color space in PNG and APNG output

[robto09]

Summary

Paparazzi writes snapshot images as PNG and APNG files whose pixels are in the sRGB color space, but the files never declared it. Decoders have to guess the color space of every snapshot. I made the color space explicit by writing the PNG sRGB chunk with the Perceptual rendering intent, as the first step toward wide color screenshots (see #40).

What I changed

• PngConstants: added the SRGB chunk header and a constant for the Perceptual rendering intent (value 0).
• ApngWriter: every PNG now carries an sRGB chunk with a one byte payload. I placed the chunk after the header chunks (IHDR, and acTL/fcTL for animated output) and before IDAT, which is where the PNG specification requires it.
• ApngReader: I taught the reader to accept the sRGB chunk and skip its payload, so files written by the updated writer round trip cleanly. Without this branch the code does not compile, because the chunk dispatcher requires exhaustive handling of every known chunk.
• CHANGELOG.md: added an entry under Unreleased.

Why a real chunk and not a workaround

I used the sRGB chunk because it is the standard PNG mechanism for declaring the color space of image data, defined in the PNG 1.2 specification. Any conforming decoder, including ImageIO and Android's BitmapFactory, understands it. This keeps every snapshot self-describing and gives us a solid baseline when we add wide color gamut support later.

How I tested it

• Added ApngWriterTest.writesSRGBColorSpaceChunkAfterIHDR, which asserts the chunk id, its position between IHDR and IDAT, the Perceptual intent byte, and a valid CRC.
• Added ApngReaderTest.decodesPNGWithSRGBColorSpaceChunk, which writes a PNG with the updated writer and decodes it with the reader.
• Updated the four existing chunk sequence tests to include the new chunk.
• APNG suite: 20 tests, 0 failures.
• Full :paparazzi:test suite: 126 tests, 0 failures.
• I also validated the output independently: a Python PNG chunk parser reports IHDR, sRGB, IDAT, IEND with intent byte 0 and valid CRCs, and macOS sips decodes the file.

Follow ups

• The delta-*.png diagnostics written via ImageIO.write in ImageUtils are a separate output path and I left them unchanged; they can get the same treatment later.
• Third party PNGs with unknown ancillary chunks (gAMA, PLTE, tEXt) still fail in ApngReader because Header.valueOf throws for chunk types outside the enum; a graceful skip for unknown chunks would be a good follow up.

Fixes #40