Daily run failure: Unit tests

[github-actions]

The daily unit tests failed. See https://github.com/Quantco/glum/actions/runs/8578861406 for details.

[stanmart]

I believe the reason for the nightly failure is this deprecation came into effect in pandas 3.0.0dev0:

[MatthiasSchmidtblaicherQC]

I think I managed to reproduce it:

import pandas as pd

pd.SparseDtype(bool, fill_value=0)

after installing the nightly version of pandas raises ValueError: fill_value must be a valid value for the SparseDtype.subtype. See also changes to test_array here. So fill_value in https://github.com/Quantco/tabmat/blob/0e22f477e134cf119789eab8279f5abf2b5b193a/src/tabmat/constructor.py#L111 should be changed to False for Booleans.

[stanmart]

@MatthiasSchmidtblaicherQC now that your PR is merged this can be closed, right?

[lbittarello]

The daily tests are still failing, but for different reasons: https://github.com/Quantco/glum/actions/runs/9153751283/job/25163209834

[stanmart]

Oh I see. #792 should fix the test_formula failures. The rest are due to scipy removing the .A shorthand for toarray(), and need to be fixed in tabmat.

[MarcAntoineSchmidtQC]

fixed with tabmat #361