BUG fix slogdet bug

[BorisMuzellec]

What does your PR implement? Be specific.

This PR fixes bugs when using pydeseq2 with pandas 2.0.

Code to reproduce the issue

First bug

counts_df = load_example_data(
    modality="raw_counts",
    dataset="synthetic",
    debug=False,
)

clinical_df = load_example_data(
    modality="clinical",
    dataset="synthetic",
    debug=False,
)


dds = DeseqDataSet(
    counts=counts_df,
    clinical=clinical_df,
    refit_cooks=True,
    design_factors="condition",
)

dds.deseq2()

would output

Fitting size factors...
... done in 0.00 seconds.

Fitting dispersions...

UFuncTypeError: Cannot cast ufunc 'slogdet' input from dtype('O') to dtype('float64') with casting rule 'same_kind'

Second bug

After fixing the first bug, DeseqDataSet.refit() would throw anndata.ImplicitModificationWarnings, making the tests fail.

Fix

The first bug is caused by pydeseq2.utils.build_design_matrix outputting boolean columns with pandas API changes. This PR fixes it by casting the design matrix to int, which is compatible with older versions of pandas.

The second bug arrises when copying anndata objects with empty indexes. This PR fixes it by checking that indexes are non-empty prior to copying.

[maikia]

LGTM, thanks

[maikia]

remove?

[BorisMuzellec]

Indeed, you caught me red-handed doing print debugging :)