[D] Scikit-Learn fixed its F-1 score calculator; you should update now

Scikit-Learn 1.3.x had a bug in its F-1 score calculator that was fixed in the latest version (1.4.0, released last week) which could produce the wrong score when the `zero_division` parameter was set to `0.0` or `np.nan`, e.g.:

>>> sklearn.__version__
'1.3.2'
>>> sklearn.metrics.f1_score(y_true=[0, 0, 1, 2, 3], y_pred=[0, 1, 0, 2, 3], zero_division=1.0, average="macro")
0.875 # Wrong

vs. (the exact same input)

>>> sklearn.__version__
'1.4.0'
>>> sklearn.metrics.f1_score(y_true=[0, 0, 1, 2, 3], y_pred=[0, 1, 0, 2, 3], zero_division=1.0, average="macro")
0.625 # …

calculator learn machinelearning metrics nan scikit scikit-learn set sklearn update