BUG/ENH: groupby quantile arraylike fails with integer columns

[fujiaxiang]

When columns are integers, df.groupby(label).quantile(<arraylike>) fails.
This PR fixes this issue by telling MultiIndex to only look for positional levels during an reordering operation.
Two new internal methods in MultiIndex class, and one new parameter for DataFrame.reorder_levels and MultiIndex.reorder_levels are added during the debugging process.

• closes groupby.quantile(<arraylike>) fails with AssertionError #30289
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[jreback]

this is not the right solution as we are not adding a confusing at best parameter to a public function

this entire routine needs to be refactored to avoid the current method which is highly suspect now

[fujiaxiang]

this is not the right solution as we are not adding a confusing at best parameter to a public function

this entire routine needs to be refactored to avoid the current method which is highly suspect now

hi Jeff, thanks for review.
Now that I think about it, I agree with you this solution is too rough.
The idea of positional parameter came from the discussions in the issue (#30289).
By "this entire routine needs to be refactored" I assume you mean GroupBy.quantile should not rely on DataFrame.reorder_levels at all? Could you let me know a bit more details about the design you are thinking of?
Anyway, should I close this PR and start another one when ready? Or do you think maybe some of it is useful and I should keep this?

[jreback]

group by quantile needs the refactor to avoid the need to reindex the multiindex which is causing the issue now)

you can close or push to this PR up to u

[fujiaxiang]

Sorry I messed up the different commits and reverts in this branch. Will close this PR and open a new one.