chore: upgrade to spec0

[ilan-gold]

See https://scientific-python.org/specs/spec-0000/ for more info. The concerning part is

which means we can't have both 1.13 1.12 scipy and 2.0 numpy simultaneously 1.13 1.12 scipy doesn't support 2.0 numpy.

Few options:

1. Manually "edit" min job to either downgrade numpy or upgrade scipy if we change the pyproject.toml to reflect spec0 properly
2. Add more min jobs that have different dependency chains (i.e., do both of the above separately)
3. Leave as-is with the non spec0 deps in the pyproject.toml

• Closes #
• Tests added
• Release note added (or unnecessary)

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.59%. Comparing base (3bca27d) to head (c2af8bf).
⚠️ Report is 1 commits behind head on main.
✅ All tests successful. No failed tests found.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2143      +/-   ##
==========================================
+ Coverage   84.02%   84.59%   +0.56%     
==========================================
  Files          46       46              
  Lines        7113     7106       -7     
==========================================
+ Hits         5977     6011      +34     
+ Misses       1136     1095      -41     

Files with missing lines	Coverage Δ
src/anndata/_core/aligned_mapping.py	95.02% <100.00%> (-0.09%)	⬇️
src/anndata/_core/anndata.py	83.07% <100.00%> (-0.11%)	⬇️
src/anndata/_core/extensions.py	96.15% <100.00%> (-0.10%)	⬇️
src/anndata/_core/merge.py	85.09% <100.00%> (+1.24%)	⬆️
src/anndata/_io/h5ad.py	93.62% <100.00%> (-0.04%)	⬇️
src/anndata/_io/read.py	52.55% <100.00%> (-29.41%)	⬇️
src/anndata/_io/specs/lazy_methods.py	96.15% <100.00%> (ø)
src/anndata/_io/specs/registry.py	94.55% <100.00%> (+0.05%)	⬆️
src/anndata/_io/write.py	53.68% <100.00%> (-22.66%)	⬇️
src/anndata/_io/zarr.py	82.02% <100.00%> (-0.20%)	⬇️
... and 11 more

... and 7 files with indirect coverage changes

[flying-sheep]

ah, this was the issue: dda018f

we have multiple registries, this can’t be a class variable.

[flying-sheep]

regarding numpy 2 vs scipy 1.13, it’s a weird situation.

we can’t drop numpy 1 jobs (and upgrade to numpy-2-only APIs) while supporting scipy 1.13, so effectively we’d test with numpy 1 until then. I guess SPEC 0 support therefore means “if you depend on scipy, you gotta support numpy 1 until scipy 1.13 support is dropped”.

I guess I have to re-introduce numpy 1 support in fast-array-utils since it supports scipy.