ENH: Add engine keyword to expanding.apply to utilize Numba

[mroeschke]

• closes ENH: Add engine keyword argument to expanding.apply to leverage Numba #30936
• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

[WillAyd]

Thanks! Do we have any benchmarks for this?

[WillAyd]

Is this totally equivalent? I get the impression previously should have been *args and **kwargs so a little strange as is, but I don't think converting to None is totally the same?

[mroeschke]

Previously the docs described rolling/expanding.apply being able to take *args, **kwargs: https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.core.window.Expanding.apply.html

But in actuality, the arguments were args=(), kwargs={}: https://github.com/pandas-dev/pandas/blob/0.25.x/pandas/core/window.py#L2128

Since the docs were incorrect, and the prior arguments has a mutable default argument, I changed rolling.apply to be args=None, kwargs=None (later converting to args=(), kwargs={} internally if still None), and I am doing the same here for expanding.apply.

So slightly changing the behavior, but taking advantage of the fact the docs were incorrect in the first place.

[WillAyd]

Can you add annotation for this? I assume Dict[str, Any] unless can further subset the provided values

[jreback]

lgtm.

[jreback]

actually ok for 1.0.0; can you add this issues / PR number to the whatsnew note that already existis for rolling.apply