fix obscure pandas bug

[austinjpaul]

Requesting historical data for August 2018, e.g. would not work because pandas decided to return an Index instead of a DateTimeIndex.
This should fix that by manually coercing it.

>>> cal = USFederalHolidayCalendar()
>>> cal.holidays(start=datetime.date(2021,8,1), end=datetime.date(2021,8,31))
DatetimeIndex([], dtype='datetime64[ns]', freq=None)
>>> cal.holidays(start=datetime.date(2020,8,1), end=datetime.date(2020,8,31))
DatetimeIndex([], dtype='datetime64[ns]', freq=None)
>>> cal.holidays(start=datetime.date(2019,8,1), end=datetime.date(2019,8,31))
DatetimeIndex([], dtype='datetime64[ns]', freq=None)
>>> cal.holidays(start=datetime.date(2018,8,1), end=datetime.date(2018,8,31))
Index([], dtype='object')

[friendchris]

I love this bug and I love finding it.

[austinjpaul]

Filed with pandas: pandas-dev/pandas#49075

[roadswitcher]

I love this bug and I love finding it.

It was hilarious fun to walk thru the pandas codebase to find this.

pandas PR should be merged soonish.