Comparing against datetime.datetime and pd.Timestamp

Not sure if exactly bug and what performance implications there are but it'd be more user friendly if supported:

1.) comparing against datetime

import datetime
import pandas as pd
import xarray as xr
ds = xr.Dataset(coords={'int': [0, 1, 2]})
ds['data'] = ('int', [0, 5, 6])
ds.coords['time'] = ('int', pd.date_range('2017-02-01', '2017-02-03'))
ds = ds.where(ds['time'] > datetime.datetime(2017, 2, 2))
ds

TypeError: '>' not supported between instances of 'int' and 'datetime.datetime'

2.) pd.Timestamp

import datetime
import pandas as pd
import xarray as xr
ds = xr.Dataset(coords={'int': [0, 1, 2]})
ds['data'] = ('int', [0, 5, 6])
ds.coords['time'] = ('int', pd.date_range('2017-02-01', '2017-02-03'))
ds = ds.where(ds['time'] > pd.to_datetime('2017-02-02'))
ds

This works though when converting to np.datetime64

import datetime
import pandas as pd
import xarray as xr
ds = xr.Dataset(coords={'int': [0, 1, 2]})
ds['data'] = ('int', [0, 5, 6])
ds.coords['time'] = ('int', pd.date_range('2017-02-01', '2017-02-03'))
ds = ds.where(ds['time'] > pd.to_datetime(['2017-02-02']).values)
ds

Start by running the two datetime comparison examples and the working np.datetime64 example from the issue. Trace the comparison and where-clause path used by ds['time'] > the scalar values. Done means datetime.datetime and pd.Timestamp comparisons work as shown without breaking the existing np.datetime64 behavior.