expect_column_values_to_be_in_set fails on datetimes

[RoyalTS]

This seems like it should validate correctly but doesn't:

import pandas as pd
import great_expectations as ge
 
df = pd.DataFrame({'a': pd.to_datetime(['2018-01-01'])})
df_ge = ge.from_pandas(df)
df_ge.expect_column_values_to_be_in_set('a', pd.to_datetime(['2018-01-01']))

[jcampbell]

The issue here is that GE is converting your values set to string prior to saving the expectation (to ensure the expectation configuration objects are fully json serializable), so the resulting object fails the comparison to your datetime object in the dataframe.

Because datetimes are such an important case, for several other expectations (e.g. expect_column_values_to_be_between), we specifically allow you to set a flag parse_strings_as_datetimes to enable this type of comparison; that should clearly be available in this case.

[jcampbell]

The new PR addresses the acute pain of this issue, but it's still counterintuitive to have to specify parse_strings_as_datetimes when none of the types you're passing in are strings--it's accommodating the need for the generic storage format of the expectation configuration.

I think this touches on the need for types #207 and so if @RoyalTS agrees suggest we close this once we merge the PR as an acute fix and follow this issue there.