Fix diffmap plotting

[Koncopd]

Fixes https://github.com/scverse/scanpy/issues/2254 https://github.com/scverse/scanpy/issues/2227

Before there was a hard coded offset for the diffmap basis (https://github.com/scverse/scanpy/blob/8736fc3b02a6af3b7214481c2cf78a3f8f5e06bd/scanpy/plotting/_tools/scatterplots.py#L1077) and later it disappeared. This restores the offset for diffmap, and also provides an option in embedding to select basis columns before components, dimensions etc.

[Zethson]

@Koncopd can you get the CI to pass, please? Happy to review then.

[Koncopd]

I am not sure, what happens with CI. I don't think it is because of the changes in this PR.

[ivirshup]

/azp run

[azure-pipelines[bot]]

Azure Pipelines successfully started running 1 pipeline(s).

[ivirshup]

@Koncopd, there was a bugged release of pip. I think this should work now that there has been a bugfix

[codecov[bot]]

Codecov Report

Merging #2255 (9759293) into master (8e5fedb) will decrease coverage by 0.00%. The diff coverage is 75.00%.

@@            Coverage Diff             @@
##           master    #2255      +/-   ##
==========================================
- Coverage   71.82%   71.82%   -0.01%     
==========================================
  Files          98       98              
  Lines       11539    11542       +3     
==========================================
+ Hits         8288     8290       +2     
- Misses       3251     3252       +1     

Impacted Files	Coverage Δ
scanpy/plotting/_tools/scatterplots.py	88.09% <75.00%> (-0.16%)	:arrow_down:

[Koncopd]

Ok, i will merge and check, thx.

[Koncopd]

@giovp Could you check why the visium test fails? I don't think it is related to this PR.

[giovp]

ah it seems a really minor tolerance thingy, try to increase it to 16 should be fine

save_and_compare_images = image_comparer(ROOT, FIGS, tol=15)

[Koncopd]

/azp run

[azure-pipelines[bot]]

Azure Pipelines successfully started running 1 pipeline(s).

[dburkhardt]

@ivirshup is this the accepted solution for https://github.com/scverse/scanpy/pull/2280?

Just trying to figure out if we can close the PR and fix the bug.