FIX warning using polars DataFrames in DecisionBoundaryDisplay.from_estimator

[patrickkwang]

Reference Issues/PRs

Fixes #28717, which is analogous to #23311, but for polars.

See also #25896.

What does this implement/fix? Explain your changes.

This fix is analogous to #23318. We identify when the predictors passed to DecisionBoundaryPlot.from_estimator is a polars DataFrame and refrain from converting it to a numpy array and stripping the column names.

Any other comments?

The way we do this is a bit awkward because we want to avoid importing polars. A nicer way to do this would be something like the following:

X_grid = pl.DataFrame({
    column: series
    for column, series in zip(X.columns, (xx0.ravel(), xx1.ravel()))
})

... but that would require import polars as pl.

Instead we drop the existing columns and add columns with the same names and the data from meshgrid.

[github-actions]

✔️ Linting Passed

All linting checks passed. Your pull request is in excellent shape! ☀️

_{Generated for commit: b66a38d. Link to the linter CI: here}

[jeremiedbb]

Thanks for the PR @patrickkwang. ping @glemaitre who made the fix for pandas dataframes for another review.

[jeremiedbb]

Please add an entry in the v1.4.rst change log. Otherwise LGTM. Thanks @patrickkwang.

[patrickkwang]

Please add an entry in the v1.4.rst change log. Otherwise LGTM. Thanks @patrickkwang.

Done.

[jeremiedbb]

We've decided that the 1.4.2 release will only be about numpy 2 support, so I moved the changelog entry into v1.5.rst, which is expected in a month. Sorry for the misleading request.

[glemaitre]

Otherwise LGTM.

[jeremiedbb]

Thanks @patrickkwang