BUG: round-trip datetime-period-datetime not preserving NaT

[jreback]

In [76]: data
Out[76]: 
0   2014-10-01
1   2014-10-01
2   2014-10-31
3   2014-11-15
4   2014-11-30
5          NaT
6   2014-12-01
dtype: datetime64[ns]

In [78]: DatetimeIndex(data).to_period(freq='M')
Out[78]: 
<class 'pandas.tseries.period.PeriodIndex'>
freq: M
[2014-10, ..., 2014-12]
length: 7

In [79]: DatetimeIndex(data).to_period(freq='M').values
Out[79]: array([ 537,  537,  537,  538,  538, 3507,  539])

In [80]: DatetimeIndex(data).to_period(freq='M').to_timestamp()
Out[80]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2014-10-01, ..., 2014-12-01]
Length: 7, Freq: None, Timezone: None

In [81]: DatetimeIndex(data).to_period(freq='M').to_timestamp().values
Out[81]: 
array(['2014-09-30T20:00:00.000000000-0400',
       '2014-09-30T20:00:00.000000000-0400',
       '2014-09-30T20:00:00.000000000-0400',
       '2014-10-31T20:00:00.000000000-0400',
       '2014-10-31T20:00:00.000000000-0400',
       '2262-03-31T20:00:00.000000000-0400',
       '2014-11-30T19:00:00.000000000-0500'], dtype='datetime64[ns]')