Release candidate 1 of lcms2-2.20

lcms2-2.20 is a maintenance release

Changes:

• Added a plug-in type for profile header parsing.
• Fixed premultiplied alpha on planar 16 bit and optimized matrix-shaper transforms.
• Fixed planar half float and Lab double formatters, and planar strides in fast_float.
• Fixed gamut checking with two profiles and a possible crash with malformed profiles.
• Recovered neutrality correction in black point detection.
• Made cmsCreateLab2Profile white point handling consistent with cmsCreateLab4Profile.
• Fixed UTF8 buffer size calculation.
• CGATS parser is now more permissive on certain properties.
• Fixed callback type mismatch in resampled CLUT evaluation.
• Fixed a crash in PostScript CSA generation.
• Fixed memory leaks in black preserving intents and .cube loading.
• cmsChangeBuffersFormat now correctly rejects format changes on floating point transforms.