Fix glum issue-872

[mlondschien]

but this fixes Quantco/glum#872 for me. I still get a warning

Iteration 99:   8%|████████████▎                                                                                                                                     | 0.42/5.0 [0.07s/it, gradient norm=3.834443702192064]
/Users/mlondschien/mambaforge/envs/icu-experiments/lib/python3.10/site-packages/glum/_solvers.py:336: ConvergenceWarning: IRLS failed to converge. Increase the maximum number of iterations max_iter (currently 100)
  warnings.warn(

[stanmart]

Interesting, probably some float underflow/underflow? Not sure tbh, but the fix looks good, and if it works... :)