Skip labels before loop

[johnarevalo]

When there are many labels to be skipped, having this validation inside the for loop makes kbet computation significantly slow:

scib/scib/metrics/kbet.py

Lines 114 to 122 in ed3e284

	# subset by label
	adata_sub = adata_tmp[adata_tmp.obs[label_key] == clus, :].copy()
	# check if neighborhood size too small or only one batch in subset
	if np.logical_or(
	adata_sub.n_obs < 10, len(adata_sub.obs[batch_key].cat.categories) == 1
	):
	print(f"{clus} consists of a single batch or is too small. Skip.")
	score = np.nan

This PR extract such checks outside the for loop to do them in a single pass.

[codecov]

Codecov Report

Attention: Patch coverage is 80.00000% with 6 lines in your changes missing coverage. Please review.

Project coverage is 54.63%. Comparing base (9271d92) to head (7d9e357).
Report is 2 commits behind head on main.

Files with missing lines	Patch %	Lines
scib/metrics/kbet.py	80.00%	6 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #394      +/-   ##
==========================================
+ Coverage   54.60%   54.63%   +0.02%     
==========================================
  Files          22       22              
  Lines        1683     1684       +1     
==========================================
+ Hits          919      920       +1     
  Misses        764      764              

Flag	Coverage Δ
unittests	54.63% <80.00%> (+0.02%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
scib/metrics/kbet.py	72.22% <80.00%> (+0.25%)	⬆️

[adamgayoso]

@johnarevalo would you be willing to make a similar PR to our scib-metrics package here? https://github.com/YosefLab/scib-metrics

[johnarevalo]

Sure! I'll port these changes to scib-metrics and tag you when the PR is ready.