REGR: missing vals not replaceable in categorical

[mzeitlin11]

• closes BUG: series.replace(np.nan,..) on categorical series does not replace #40472
• tests added / passed
• Ensure all linting tests pass, see here for how to run them
• whatsnew entry

This restores behavior to before the regression, but I think also deprecating this behavior while patching might be reasonable. Since categoricals represent all missing values the same way, this allows unintuitive behavior like

ser = pd.Series([None, "b"]).astype('category')
ser = ser.replace({np.nan: 'c'})
print(ser)

replacing the missing value even though None and np.nan are different missing values. Forcing usage of the more explicit fillna seems more reasonable in this case.

[pep8speaks]

Hello @mzeitlin11! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

There are currently no PEP 8 issues detected in this Pull Request. Cheers! 🍻

Comment last updated at 2021-03-27 17:07:49 UTC

[jreback]

this is a special case: #40472 (comment)

let's discuss first

[mzeitlin11]

Ah makes sense, will close until discussion resolved