Update for scikit-learn 1.8.0 and pandas 3.0.0 compatibility

[rcmdnk]

scikit-learn 1.8.0

• Reference:
  https://scikit-learn.org/stable/whats_new/v1.8.html
  DOC improve doc for _check_n_features and _check_feature_names and fix their usages scikit-learn/scikit-learn#31585

In scikit-learn 1.8.0, the private _check_input method was removed from
FunctionTransformer. Instead, validate_data is now called directly inside
FunctionTransformer.fit.

To align with this change, this PR updates:

• mne_features/feature_extraction.py
  • FeatureFunctionTransformer.fit

so that it calls validate_data directly, mirroring the behavior of
FunctionTransformer.fit in scikit-learn 1.8.0.

This avoids reliance on removed private APIs and ensures forward compatibility.

Note: scikit-learn 1.8.0 was released for Python >= 3.11.

---

pandas 3.0.0

• Reference:
  https://pandas.pydata.org/docs/dev/whatsnew/v3.0.0.html

In pandas 3.0.0, calling X.values on a string-typed column now returns a
pandas.arrays.StringArray instead of a NumPy ndarray.

As a result, numpy.testing.assert_equal no longer works as expected and raises:

ValueError: The truth value of an array with more than one element is ambiguous.
Use a.any() or a.all().

To ensure correct and stable comparisons across pandas versions, this PR updates
the relevant tests in:

• mne_features/tests/test_univariate.py

Specifically, occurrences of X.values are replaced with X.to_list(), which
provides a plain Python list and avoids ambiguity when comparing string data.

[bruAristimunha]

ping @larsoner and @agramfort

[larsoner]

This is going to lead to an import issue on any version of sklearn that doesn't have this, right? I think over in MNE-Python I figured out this meant 1.6+

https://github.com/mne-tools/mne-python/blob/87d1051d6932383e825ba377b21a221b2670db11/mne/decoding/_fixes.py#L6-L7

Can you add a corresponding pin to >=1.6 to

mne-features/setup.py

Line 63 in 331ca47

install_requires=['numpy', 'scipy', 'numba', 'scikit-learn', 'mne',

[rcmdnk]

@larsoner
Thanks for pointing this out.
I've now updated requirements in environment.yml/setup.py to scikit-learn>=1.6.