(fix): cache arrays in BaseCompressedSparseDataset

[ilan-gold]

• Noticed while creating the screenshot tutorial for the announcement
• Tests added
• Release note added (or unnecessary)

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 84.55%. Comparing base (d61e09c) to head (1dcf7ad).

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1744      +/-   ##
==========================================
- Coverage   87.01%   84.55%   -2.46%     
==========================================
  Files          40       40              
  Lines        6059     6075      +16     
==========================================
- Hits         5272     5137     -135     
- Misses        787      938     +151     

Files with missing lines	Coverage Δ
src/anndata/_core/sparse_dataset.py	93.50% <100.00%> (+0.48%)	⬆️
src/anndata/_io/specs/lazy_methods.py	100.00% <100.00%> (ø)

... and 8 files with indirect coverage changes

---- 🚨 Try these New Features:

• Flaky Tests Detection - Detect and resolve failed and flaky tests

[flying-sheep]

Does what it promises and has nice tests.

However I think Isaac only cached the longest array intentionally. I forgot the reason though.

Sure, we now have much bigger data so maybe if there was a tradeoff, things are now different. But maybe worth checking.

[ilan-gold]

However I think Isaac only cached the longest array intentionally. I forgot the reason though.

We only cached indptr previously as a np.ndarray i.e., we read it into memory because it needs to be read in anywya ever time you make an access.

As for the other arrays, I don't think there was a good reason - I think we wanted to just start with that but I posted on zulip.

[ilan-gold]

(I was the one who started doing the caching)

[flying-sheep]

Hm, since these arrays are storage classes anyway, and probably don’t cost any noticable amount of memory, I don‘t foresee a problem.

[ilan-gold]

Let's review this in the absence of an answer

[ilan-gold]

(Or rather, give approval since you already seem to have reviewed)