Rolling on DataFrameGroupBy duplicated index column when part of the grouping cols is from index

[phofl]

…grouping cols is from index

• closes BUG: Series.groupby.rolling duplicates index when grouping over index #36794
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Could not find a elegant way, to avoid adding the components in the for loop, so I dropped it from the resulting MultiIndex again.

Had to change a test, which depended on the wrong behavior.

cc @mroeschke

[jreback]

this is way to special cases; we need to figure out why this is happening and adjust the logic more generally.

[phofl]

This is happening because of

        for key, values in self._groupby.grouper.indices.items():
            for value in values:
                data = [
                    *com.maybe_make_list(key),
                    *com.maybe_make_list(grouped_object_index[value]),
                ]
                result_index_data.append(tuple(data))

        result_index = MultiIndex.from_tuples(
            result_index_data, names=result_index_names
        )

which is a few lines below. self._groupby.grouper.indices.items()represents the grouping columns and grouped_object_index[value] the index used for the rolling operation. If parts of the grouping columns are from the index, we duplicate these here. As I wrote, I could not figure out a simple way to avoid this within the loop.

[mroeschke]

As explained in #36794 (comment), I don't view this as a bug that needs fixing