Add support for reading TIFFs with PlanarConfiguration=2

[kkopachev]

Fixes #4621 .

Adding support for reading of compressed TIFFs with planar config = separate.

Since TIFF reading is done in tiles or strips (might be multiple rows), I've added reading it by sample and shuffling into image data. For 16bit images I had to create new single channel unpackers.

There is also special handling for RGBa images with separate plane storage, since unpacking it sample by sample would leave data in RGBa, so there is extra step required to convert it to RGBA.

I also thought of using line-interleaved unpackers, but they seem to be no use since tiles smaller than width of image.

[kkopachev]

Thanks @radarhere for lint fixes.

Not sure what's going on with MacOS builds and why they are failing. I don't have Mac in possession to test it first hand, so I tried enabling TRACE logs for TiffDecode and pushed them into separate branch under my form to see in GA where it crashes. Surprisingly it finished with no errors. So I don't have any ideas where to look.
Any help appreciated.

[radarhere]

I've created PR kkopachev#2 to resolve the macOS exception

[kkopachev]

@radarhere Thank you for update, looks cleaner. I was trying to use higher-level language approach to swap pointer :)
I'm curious why original solution did not work. Any insights why it was failing?

[radarhere]

No insights. If it was a simple mistake, then you would think it would be failing for more platforms than just macOS. I just found that ((*unpacker)[plane]) was where the error was being triggered, so I started trying alternatives and somewhere in there thought to remove the need for dereferencing entirely. So some of it was just trial and error using my macOS machine, some of it was using trial and error on GitHub Actions.

[wiredfool]

I want to review this before merge for safety. Not that I’ll catch everything but I want to make a pass on it.

[kkopachev]

I've added tests for new unpackers. Travis failure is unrelated.

[kkopachev]

Any updates here? Would love this to make it into release.

[hugovk]

@wiredfool Do you think you'll have time to check this before release day?

[wiredfool]

Apparently not. It was on my list for last weekend, but I didn’t get to it.

[imfantuan]

Hey guys, any updates here?
@hugovk @kkopachev @wiredfool

[aclark4life]

@imfantuan Needs a rebase at least

[kkopachev]

Rebased.
I had to skip added tests for big endian system, following existing pattern.

[wiredfool]

This has major conflicts with a patch for a CVE that will be released in the next version.

[wiredfool]

See #5178

[hugovk]

Rebased in #5178.