Fix sort_values with Timestamp data

[jrbourbeau]

• Closes sort_values() can fail with Timestamp data #9641
• Tests added / passed
• Passes pre-commit run --all-files

[rjzamora]

Thanks @jrbourbeau !

It's interesting that the motivating bug is caused by the use of a single-column DataFrame instead of a Series. Perhaps we should raise an error in set_partitions_pre when s is not series-like?

[jrbourbeau]

Thanks for the review @rjzamora. Running the test suite with a DataFrame-like check added locally, it doesn't look like there are other time where s is a DataFrame (at least as measured by our existing test coverage). FWIW I think the reason we didn't catch this earlier is because pandas has inconsistent behavior when handling DataFrames as input to searchsorted -- I've opened an upstream issue here pandas-dev/pandas#49620. We certainly can add a Series check here, but it might be overkill