IO refactor

[ilia-kats]

Switch to AnnData's public API as much as possible.

I was hoping to be able to get rid of most of our custom implementations and just use read_elem for everything, but unfortunately that is not possible:

• for h5ad files, read_elem does not handle backed mode and AnnData's read_h5ad only supports file paths, not opened files/groups.
• for Zarr files, we have the write_data argument which is not supported by AnnData. This was presumably introduced as an analog to _write_h5mu's write_data. However, the _write_h5mu write_data argument is required for handling backed h5mu files in MuData.write_h5mu. Zarr files cannot be backed, so that argument is presumably not needed and can be removed.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 91.47%. Comparing base (5ccf921) to head (da9e2d3).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #133      +/-   ##
==========================================
+ Coverage   90.98%   91.47%   +0.48%     
==========================================
  Files          11       10       -1     
  Lines        1797     1759      -38     
==========================================
- Hits         1635     1609      -26     
+ Misses        162      150      -12     

Files with missing lines	Coverage Δ
src/mudata/_core/io.py	94.96% <100.00%> (-0.02%)	⬇️
src/mudata/_core/mudata.py	93.12% <100.00%> (-0.01%)	⬇️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.