BUG/CLN: datetime64/timedelta64

[jreback]

Various bugs related to datetime64s

• Properly convert np.datetime64 objects in a Series, BUG: conversion from np.datetime64 has odd numpy buglet #3416

This would convert to object dtype previously

In [1]: dates = [
   ...:     np.datetime64(datetime.date(2013, 1, 1)),
   ...:     np.datetime64(datetime.date(2013, 1, 2)),
   ...:     np.datetime64(datetime.date(2013, 1, 3)),
   ...: ]

In [2]: s = pd.Series(dates)

In [3]: s
Out[3]: 
0   2013-01-01 00:00:00
1   2013-01-02 00:00:00
2   2013-01-03 00:00:00
dtype: datetime64[ns]

• Do not allow astypes on datetime64[ns] except to object, and
  timedelta64[ns] to object/int, BUG: what to do with astype on datetime64 #3425
• Do not allow datetimelike/timedeltalike creation except with valid types
  (e.g. cannot pass datetime64[ms], CLN: convert_objects(convert_dates='force') should use to_datetime() #3423

Any creation/astype of a datetimelike to a non accepted dtype will raise

In [6]: Series([Timestamp('20130101'),Timestamp('20130102')],dtype='datetime64[s]')
TypeError: cannot convert datetimelike to dtype [datetime64[s]]

[jreback]

@wesm @y-p These are various datetime fixes, really minor. They basically disallow datetime operations that can cause weird issues, e.g. trying to convert a datetime64[ns] to something which we don't support now will raise