Pandas COMPAT: force int when doing unit conversions in pd.to_datetime

from SO xref to #11758

We should force int conversions when unit is passed (as by definition these are int-like things).

In [21]: s = pd.Series(["1440643875", "1440644191", "1440645638", "1440998720"], dtype=object)

In [22]: pd.to_datetime(s,unit='s')
ValueError: year is out of range

In [23]: pd.to_datetime(s.astype(int),unit='s')
Out[23]: 
0   2015-08-27 02:51:15
1   2015-08-27 02:56:31
2   2015-08-27 03:20:38
3   2015-08-31 05:25:20
dtype: datetime64[ns]

Comment From: gliptak

This seem to have been corrected in current HEAD:

In [5]: s = pd.Series(["1440643875", "1440644191", "1440645638", "1440998720"], dtype=object)

In [6]: pd.to_datetime(s,unit='s')
Out[6]: 
0   2015-08-27 02:51:15
1   2015-08-27 02:56:31
2   2015-08-27 03:20:38
3   2015-08-31 05:25:20
dtype: datetime64[ns]

Comment From: jreback

this was closed by #13003 and is already sufficiently tested.