Possible additional common fixes

[noizuy]

Obviously, there are a ton of other fixes that might be better to hardcode tests for instead of hoping matchcolors finds them. Here are some I've heard of but haven't seen often enough to necessarily warrant hardcoded tests:

• Luma/chroma-only truncation
• Chroma truncation with neutrals untouched
• Truncation from different bit depths; 12, 14, and even 11 seem surprisingly common
• Strong green tints like offsets of -1 or -2 (8-bit) on R and B
• Chroma truncation from signed (float), implying differing signs above and below neutrals
• Combinations of common errors, e.g. truncation followed by redundant TV to PC level conversion or the gamma bug along with BT.709 to BT.601 conversion
• Double BT.601 to BT.709 conversion
• Various fmtconv functions

Some of the truncation errors could be fixed by adding a YUV offset test in the matchcolors fixes.