BUG: DatetimeIndex with nanosecond frequency does not include end

[ChadFulton]

I'm not sure if this is a bug or intended behavior, but documentation says "If periods is none, generated index will extend to first conforming time on or just past end argument", and it appears here that in the nanosecond frequency case, the generated index only extends to just before the end argument.

Code Sample, a copy-pastable example if possible

> start = pd.Timestamp(1)
> end = pd.Timestamp(4)
> pd.DatetimeIndex(start=start, end=end, freq='N')
DatetimeIndex(['1970-01-01 00:00:00.000000001',
               '1970-01-01 00:00:00.000000002',
               '1970-01-01 00:00:00.000000003'],
              dtype='datetime64[ns]', freq='N')

whereas a similar call with annual frequency gives:

> start = pd.Timestamp('1971')
> end = pd.Timestamp('1974')
> pd.DatetimeIndex(start=start, end=end, freq='AS')
DatetimeIndex(['1971-01-01',
               '1972-01-01',
               '1973-01-01',
               '1974-01-01'],
              dtype='datetime64[ns]', freq='AS-JAN')

output of pd.show_versions()

pandas: 0.18.0

[sinhrks]

Thanks for the report. Because the range is created by np.arange, we should have +1 margin if offset is Nano.

• https://github.com/pydata/pandas/blob/master/pandas/tseries/index.py#L2015

PR is appreciated:)