Updated Backsub and docs

[kbestak]

Hi,

We've recently pushed an updated Backsub version 0.5.1 that is more lightweight, as efficient, and allows for compression of autofluorescence-subtracted pyramidal OME-TIFFs (by default with lzw). I've updated the docs accordingly.

Compression can be toggled off by setting --compression none if users do not want that functionality.

Introducing the compression, from our testing, only affects running Mesmer (which in the provided container) doesn't have imagecodecs. Using Recyze is anyway recommended when using it which mitigates the potential conflict.

Backsub now throws an error if the background does not match any marker_name value, and a warning if the background column is empty.

If the background value appears multiple times in the marker_name column (even with the strong recommentation to have unique values in the marker_name column), only the first value is used as the respective background. Closes #559

I think this PR should close #570 because the issue mentioned has no possibility to appear with the updated version.

[jmuhlich]

Why lzw as the default though? zlib compresses better on 16-bit scientific images in my experience, and tifffile can read it without imagecodecs too.

[kbestak]

We chose LZW only because it worked faster than ZLIB for writing files. I can add the compression default as zlib in this PR if preferred.

[jmuhlich]

Please do make zlib the default for mcmicro. I think a little extra time writing files is worth the smaller file sizes for reduced storage cost. Also smaller files will load faster during later analysis steps since I/O is the bottleneck, not the CPU time needed for decompression (for any storage except maybe fast local SSD). Plus we're using zlib for all other tiff compression already.

[kbestak]

Done, thanks for the feedback Jeremy!

[jmuhlich]

The documentation should also state that zlib is the default compression rather than lzw. I know the tool itself has lzw as the default but I think the mcmicro docs should reflect mcmicro's behavior.

[kbestak]

That's a good point, thanks!

[jmuhlich]

I guess the singularity test is failing just because there isn't enough space on the runner to download all the container images. 🤷