Update RunCellpose to support v3 and v4

[ErinWeisbart]

mad props to @DavidStirling at glencoesoftware for his work on v3!

[ErinWeisbart]

Need to add in functionality of if input file location is Elsewhere... then accept URLs to download models from URL (which will aid in very edge case of wanting to use RunCellpose in Distributed-CellProfiler without mounting bucket. Does still require model to be in public location).
(Will create a separate issue and not include this enhancement in this PR)

[ErinWeisbart]

Thanks to @viksyn2 for GPU test edit suggested in #273

[ErinWeisbart]

(Concurrent with merge into repo, we need to make official CellProfiler/RunCellpose Dockers for omnipose, v3, and v4)

[bethac07]

Awesomesauce!

(Concurrent with merge into repo, we need to make official CellProfiler/RunCellpose Dockers for omnipose, v3, and v4)

Are the Dockerfiles in the repo up to date, aka can we just re-tag the images you've been using for testing?

[ErinWeisbart]

Yes, those are the Dockerfiles that I used to make my own Dockers that have been testing on so they should be good to go.

[bethac07]

Also, can you check if #269 is superceded by this, and/or if we should include it? thanks!

[ErinWeisbart]

This does not supersede #269 per se, but I have not ever successfully run with GPU using Docker with any version of RunCellpose. (I marked this PR as ready to review as we have a time crunch for getting it merged but was planning on filing an issue saying that it still needed GPU-Docker testing and potentially debugging)

[bethac07]

Yes, those are the Dockerfiles that I used to make my own Dockers that have been testing on so they should be good to go.

Can you let me know the names they are currently known as, so I can just re-tag rather than rebuild, at your convenience? Ty!

but I have not ever successfully run with GPU using Docker with any version of RunCellpose.

Ah, ok, I'll take a look if I can, otherwise we'll merge what we have and revisit later, thanks.

[ErinWeisbart]

gnodar01/runcellpose_omnipose_no_pretrained:0.0.1
erinweisbart/cellpose:3.1.1.2
erinweisbart/cellpose:4.0.5

[bethac07]

@ErinWeisbart Merged, but question - any especial reason to have it be cellprofiler/cellpose:3.1.1.2 vs cellprofiler/runcellpose_with_pretrained:3.1.1.2 (or no_pretrained, whichever is correct) to follow our Cellpose2 structure? I will tag as directed for now, but let me know if we should re-tag to more parallel later

[ErinWeisbart]

Sorry, should be _with_pretrained.