References #19614.

What does this implement/fix? Explain your changes.

Gives a FutureWarning when a user configures a LibSVM learner with 'linear' kernel but also non-default gamma.

from sklearn.datasets import load_iris
from sklearn.svm import SVC

x, y = load_iris(return_X_y=True)
clf = SVC(kernel='linear', gamma=1e-6)
clf.fit(x, y)
print(clf.score(x, y))

previous output:

0.9933333333333333

current output:

...\scikit-learn\sklearn\svm\_base.py:220: FutureWarning: Setting 'gamma' when using 'linear' kernel may raise a `ValueError` starting in version 1.1 (renaming of 0.26).
  warnings.warn(
SVC(gamma=1e-06, kernel='linear')
0.9933333333333333

Any other comments?

Went with a FutureWarning instead of the DeprecationWarning as discussed in the issue, because:

• DeprecationWarning is ignored by default (unless it's code from __main__),
• It does not remove old functionality (setting gamma with kernel='linear' had no function).

I tried to find guidelines for this in the dev docs, but it either isn't there or I couldn't find it.
Just let me know if I should revert it to DeprecationWarning.
Also wasn't sure if a note of this should be added to the docs, since they already describe which parameter affects which kernel.

Tests ran with Windows/Py3.8:

pytest sklearn/svm
pytest doc/modules/svm.rst
pytest sklearn/tests/test_common.py -k SVC