BUG: min/max on empty categorical fails

[OliverHofkens]

• tests added / passed
• passes black pandas
• passes git diff upstream/master -u -- "*.py" | flake8 --diff
• whatsnew entry

Currently on master, the following fails:

import pandas as pd

# Empty Dataframe
df = pd.DataFrame({
    "cat": pd.Categorical([], categories=list("ABC"), ordered=True),
    "val": pd.Series([], dtype="int64")
})

df["cat"].max()

ValueError: zero-size array to reduction operation maximum which has no identity

Other dtypes return NaN when calling min() or max() on an empty Series. This PR changes the behavior of Categoricals to be the same.

Note that current behavior causes a downstream bug in Dask here: dask/dask#5645, so this PR could fix that bug as well.

[topper-123]

LGTM

[TomAugspurger]

Can you add a .. versionchanged to indicate that an NA value is returned for empty arrays.

[TomAugspurger]

Thanks!