Store dispersion trend coefficients as a dict so the dataset can be written to h5ad

[raashish1601]

Reference Issue or PRs

Fixes #254.

What does your PR implement? Be specific.

dds.write_h5ad() still fails after deseq2() with the current anndata (0.13.4):

IORegistryError: No method registered for writing <class 'pandas.Series'> into <class 'h5py._hl.group.Group'>

The cause is uns["trend_coeffs"] (and uns["vst_trend_coeffs"] after vst()), which are stored as a pd.Series. The varm["replaced"] part of the issue is already gone since replaced moved to var.

This stores both as a plain {"a0": ..., "a1": ...} dict, which anndata can write and read back. Indexing by "a0"/"a1" keeps working for users. dispersion_trend now accepts a dict (Series and arrays still work), and vst_transform reads the coefficients by key, because unpacking a dict would give the keys.

Added test_write_h5ad, which fits, runs VST, writes to h5ad, reads it back and checks the coefficients. It fails on main with the error above. tests/test_pydeseq2.py and tests/test_edge_cases.py pass, and ruff and mypy are clean.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 86.07%. Comparing base (2ca3b74) to head (0494798).

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #477      +/-   ##
==========================================
+ Coverage   86.05%   86.07%   +0.02%     
==========================================
  Files          15       15              
  Lines        1276     1278       +2     
==========================================
+ Hits         1098     1100       +2     
  Misses        178      178              

Files with missing lines	Coverage Δ
src/pydeseq2/dds.py	87.67% <100.00%> (+0.04%)	⬆️
src/pydeseq2/dispersions.py	96.84% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[raashish1601]

The py3.14-pre failure is unrelated to this change: every test fails when building an AnnData, because anndata reads pd.options.future.infer_string, which the pandas pre-release now deprecates with a Pandas4Warning (turned into an error by the test settings). The stable jobs pass.