merge_asof() must be able to operate with timezone-aware DatetimeIndex

[chrisaycock]

I can perform the following merge just fine:

left = pd.DataFrame({'date': pd.DatetimeIndex(start=pd.to_datetime('2016-01-02'),
                                              freq='D', periods=5),
                     'value1':np.arange(5)})
right = pd.DataFrame({'date': pd.DatetimeIndex(start=pd.to_datetime('2016-01-01'),
                                               freq='D', periods=5),
                      'value2':list("ABCDE")})
pd.merge_asof(left, right, on='date', tolerance=pd.Timedelta('1 day'))

However, adding a timezone to the DatetimeIndex doesn't work:

import pytz
left = pd.DataFrame({'date': pd.DatetimeIndex(start=pd.to_datetime('2016-01-02'),
                                              freq='D', periods=5, tz=pytz.timezone('UTC')),
                     'value1':np.arange(5)})
right = pd.DataFrame({'date': pd.DatetimeIndex(start=pd.to_datetime('2016-01-01'),
                                               freq='D', periods=5, tz=pytz.timezone('UTC')),
                      'value2':list("ABCDE")})
pd.merge_asof(left, right, on='date', tolerance=pd.Timedelta('1 day'))

I get the oddly worded

MergeError: incompatible tolerance, must be compat with type <class 'pandas.tseries.index.DatetimeIndex'>

The solution is actually very simple. _AsOfMerge._get_merge_keys() needs to check for is_datetime64tz_dtype() in addition to is_datetime64_dtype() when there is a tolerance. I should also fix the error message to be clearer.