Expose Timezone._utcoffset property for pandas compatibility

[ddelange]

This commit allows pandas to get _utcoffset from a Timezone object, taking the transision as seen from utcnow.

As explained by @sdispater in the issues linked below, pandas tries to access the Timezone._utcoffset property (which isn't currently implemented for variable Timezone objects) and subsequently tries to get Timezone.utcoffset(None) (which correctly returns None as it can't be determined without knowlede of date and time). This PR allows pandas to treat Timezones as if they were FixedTimezones (as returned for utcnow).

ref:
https://github.com/pandas-dev/pandas/blob/v1.0.5/pandas/_libs/tslibs/timezones.pyx#L152

Fixes pandas-dev/pandas#15986
Closes #131

>>> import pendulum
>>> import pandas
>>> pandas.date_range(start=pendulum.yesterday(), end=pendulum.tomorrow())
DatetimeIndex(['2020-07-15 00:00:00+02:00', '2020-07-16 00:00:00+02:00',
               '2020-07-17 00:00:00+02:00'],
              dtype='datetime64[ns, Timezone('Europe/Berlin')]', freq='D')

Pull Request Check List

• Added tests for changed code.

• Updated documentation for changed code.

Left this unticked, as it's a private property that users won't touch.

[ddelange]

Closing this as it's quite incorrect behaviour

[philsheard]

It doesn't appear that this has been addressed on either library, so if anyone wants a workaround then this worked for me:

Do a roundtrip from Pendulum > ISO format > Pandas to avoid the internal lookup that causes the warning.

converted_date = pandas.to_datetime(pendulum.today().isoformat())

Not pretty but it will clean your error logs without hurting comprehension.