Fix epoch time calculation if provided datetime has timezone

[labeneator]

I was hitting the following bug:

2016-10-07 13:26:36,709:ERROR:root:Timestamp subtraction must have the same timezones or no timezones
Traceback (most recent call last):
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/caravel/viz.py", line 328, in get_json
    data = self.json_dumps(payload)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/caravel/viz.py", line 346, in json_dumps
    return json.dumps(obj, default=utils.json_int_dttm_ser, ignore_nan=True)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/simplejson/__init__.py", line 397, in dumps
    **kw).encode(obj)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/simplejson/encoder.py", line 275, in encode
    chunks = self.iterencode(o, _one_shot=True)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/simplejson/encoder.py", line 357, in iterencode
    return _iterencode(o, 0)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/caravel/utils.py", line 380, in json_int_dttm_ser
    obj = datetime_to_epoch(obj)
  File "/Users/lmwangi/venvs/caravel/lib/python2.7/site-packages/caravel/utils.py", line 367, in datetime_to_epoch
    return (dttm - EPOCH).total_seconds() * 1000
  File "pandas/tslib.pyx", line 1071, in pandas.tslib._Timestamp.__sub__ (pandas/tslib.c:21031)
TypeError: Timestamp subtraction must have the same timezones or no timezones

This PR fixes this issue.

[bkyryliuk]

@labeneator - could you please add a test that reproduces the issue to prevent regressions?