Interleave channels when the --rgb flag is used

[melissalinkert]

This packs the RGB channels into a single tile (TIFF PlanarConfiguration 1), rather than using a separate tile per RGB channel. This has a noticeable effect on JPEG and JPEG-2000 compression for brightfield slides.

The scary-looking part of this diff is reworking how the RGB channels are iterated over. Instead of a call to writeTile for each RGB channel, all RGB channels for the current tile are read, then repacked, and then written once. For the common case where --rgb is not used (i.e. rgbChannels == 1), there should be no difference in behavior.

For the case where rgbChannels == 3, check with both default compression and --compression JPEG and/or --compression JPEG-2000. tiffdump on the output file should show that the PlanarConfiguration tag is changed with this PR; the file sizes should also be noticeably different with JPEG and JPEG-2000 compression. In local testing with CMU-1.svs, the output OME-TIFF is now approximately the same size as the original input file.

You may wish to test this together with #112, as testing this PR without the changes in #112 will result in harmless but alarming IllegalArgumentExceptions.

[sbesson]

Reporting on the initial testing of this PR using the standard CMU-1.svs sample as the input. After running bioformats2raw followed by raw2ometiff --rgb --compression JPEG, the TIFF layout matches the expectations of this PR:

(java11) sbesson@Sebastiens-MacBook-Pro-3 rgbometiff % tiffinfo ometiff/CMU-1.ome.tiff | grep Samples/P | uniq      
  Samples/Pixel: 3
(java11) sbesson@Sebastiens-MacBook-Pro-3 rgbometiff % tiffinfo ometiff/CMU-1.ome.tiff | grep "Compression S" | uniq
  Compression Scheme: JPEG
(java11) sbesson@Sebastiens-MacBook-Pro-3 rgbometiff % tiffinfo ometiff/CMU-1.ome.tiff | grep "Planar Con" | uniq   
  Planar Configuration: single image plane

Additionally the OME-TIFF size is now comparable (smaller) to the original SVS file as opposed to almost 10x larger without it

(java11) sbesson@Sebastiens-MacBook-Pro-3 rgbometiff % ls -alh samples/CMU-1.svs 
-rw-r--r--@ 1 sbesson  staff   169M  6 Nov 16:23 samples/CMU-1.svs
(java11) sbesson@Sebastiens-MacBook-Pro-3 rgbometiff % ls -alh ometiff/CMU-1.ome.tiff 
-rw-r--r--  1 sbesson  staff   133M 10 Nov 15:45 ometiff/CMU-1.ome.tiff

The code changes definitely look bigger than they really are. Hiding the whitespace diff (https://github.com/glencoesoftware/raw2ometiff/pull/113/files?diff=split&w=1) is a convenient way to review the actual changes.

I will carry on the testing next week including:

• conversion of RGB images with different parameters (tile size, compression scheme, compression quality)
• conversion of different types of RGB images (multi-RGB, multi-Z)
• conversion of non RGB images