Actually support profiles with 1D and 2D CLUTs

Parse accepts these, but transform would fail unless input_channels was
3 or 4. The other option would be to fail in parse. We're still pretty
sloppy about channel counts in data (pixel formats) vs. what the profile
expects. This is obviously a garbage image created by a fuzzer, but are
there situations where this is useful?

Bug: chromium:874433
Change-Id: Ie9672024e5d7c6b4b6f6a856f9d1454fb53a6cbb
Reviewed-on: https://skia-review.googlesource.com/147203
Reviewed-by: Mike Klein <mtklein@google.com>
Commit-Queue: Brian Osman <brianosman@google.com>
5 files changed