(fix): empty boolean mask on backed sparse matrix

[ilan-gold]

• Closes SparseDataset errors out with empty boolean mask #1320
• Tests added
• Release note added (or unnecessary)

[ilan-gold]

blocked by #1320 (comment)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (99253c7) 85.67% compared to head (5b13eb8) 83.56%.
Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1321      +/-   ##
==========================================
- Coverage   85.67%   83.56%   -2.11%     
==========================================
  Files          34       34              
  Lines        5452     5458       +6     
==========================================
- Hits         4671     4561     -110     
- Misses        781      897     +116     

Flag	Coverage Δ
gpu-tests	?

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

Files	Coverage Δ
anndata/_core/sparse_dataset.py	94.51% <100.00%> (+0.76%)	⬆️

... and 7 files with indirect coverage changes

[ivirshup]

blocked by #1320 (comment)

Do you still consider this blocked? I think it's fine to assume it acts like you'd expect a numpy array too

[ilan-gold]

blocked by #1320 (comment)

Do you still consider this blocked? I think it's fine to assume it acts like you'd expect a numpy array too

Nope, the issue is filed. You thought it was a bug too so we can't wait for a scipy release over this.

[ilan-gold]

@ivirshup I would wait for an answer from the scipy people confirming scipy/scipy#19919 (comment) before merging, just to be sure empty array creation is correct. We're so close to confirmation of future behavior that I think it's worth waiting unless we're in a rush, in which case you can merge.

[ivirshup]

I think that the array classes in scipy are going to end up returning something with .shape = (0, 0) here, but also that it's not a super big deal if we end up changing that.