Handle new oversampler nullable type incompatibility in X

[tamargrey]

closes #3974

This incompatibility only shows up when string categorical columns are present with the nullable pandas dtypes, but I chose not to add a special _handle_nullable_types implementation since it's dataset dependent vs the time series imputer that needed a different handling but that would be applied to any dataset. This is a bit of a gray area, though, so I'm open to changing it.

[codecov]

Codecov Report

Merging #4068 (19e53a2) into main (f692e0d) will increase coverage by 0.1%.
The diff coverage is 100.0%.

@@           Coverage Diff           @@
##            main   #4068     +/-   ##
=======================================
+ Coverage   99.7%   99.7%   +0.1%     
=======================================
  Files        349     349             
  Lines      37559   37583     +24     
=======================================
+ Hits       37441   37465     +24     
  Misses       118     118             

Impacted Files	Coverage Δ
...es/components/transformers/samplers/oversampler.py	100.0% <100.0%> (ø)
evalml/tests/component_tests/test_oversampler.py	100.0% <100.0%> (ø)

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

[tamargrey]

This is actually the same bug as the nullable type incompatibility, and it's why we convert categorical columns to "object" in base_sampler. There is a fix in the works scikit-learn/scikit-learn#25814 (because of the wonderful @thomasjpfan who has been a key contributor to our ability to support nullable types so widely), so I wanted to add this xfailing test to remind us to also remove that logic at the same time.

[tamargrey]

@eccabay converting caegory to object was a bug fix of yours from a while back, so I just want to make sure you get a chance to take a look at this test in case I'm missing something here.

[tamargrey]

Note - this doesn't need to get in for the upcoming release

[chukarsten]

Seems like a solid approach to me. Thanks for being so thorough in the documentation of the tests and what we need to undo when it succeeds.