Improve error message in Series.any and Series.all, when bool_only is passed and not level

[moink]

When working on xref #30999 I came across TestSeriesReductions.test_all_any_params in pandas/tests/reductions/test_reductions.py

It showed that calling the following code:

s = Series([False, False, True, True, False, True], index=[0, 0, 1, 1, 2, 2])
s.any(bool_only=True)

would raise a NotImplementedError with the error message "Series.any does not implement numeric_only", and equivalent for calling s.all the same way. This error message is confusing because the user didn't pass numeric_only, they passed bool_only.

So the messages need to change to "Series.any does not implement bool_only" and "Series.all does not implement bool_only".

what actually needs to happen is we pass bool_only thru, but this increases the scope of this PR greatly.
-- @jreback #38720

[theodorju]

Hi, I'd like to work on this one since there is no one assigned.
Should only the string be changed, or the method's argument as well from numeric_only to boolean_only?

I'm referring to method _reduce on series.py.

[weikhor]

Pandas version I used:

1.5.0.dev0+725.g9f24918759

With this pandas version, the result is worked as expected.

1. s = Series([False, False, True, True, False, True], index=[0, 0, 1, 1, 2, 2])
   s.all(bool_only=True)
   

   NotImplementedError: Series.all does not implement bool_only.
   

2. s = Series([False, False, True, True, False, True], index=[0, 0, 1, 1, 2, 2])
   s.any(bool_only=True)
   

   NotImplementedError: Series.any does not implement bool_only.
   

[rhshadrach]

Thanks @weikhor; this is tested for in test_all_any_params. Closing.