DateTimeLine does not accept pandas timestamps

I have charts that worked well with pygal v 2.1.1, but fail after an upgrade to 2.3.0.
I pass pandas Timestamps (UTC localized) as x, when I pass datetime objects everything is fine.

The exception:

```
chart = datetimeline.render(is_unicode=True)

File "/usr/local/lib/python2.7/dist-packages/pygal/graph/public.py", line 52, in render

self.setup(**kwargs)

File "/usr/local/lib/python2.7/dist-packages/pygal/graph/base.py", line 189, in setup

[rs for rs in self.raw_series if not rs[1].get('secondary')]) or []

File "/usr/local/lib/python2.7/dist-packages/pygal/graph/base.py", line 160, in prepare_values

self._x_adapt(value[0]),

File "/usr/local/lib/python2.7/dist-packages/pygal/util.py", line 311, in

fun = lambda *args, **kwargs: f(g(*args, **kwargs))

File "/usr/local/lib/python2.7/dist-packages/pygal/graph/time.py", line 35, in datetime_to_timestamp

return timestamp(x)

File "/usr/local/lib/python2.7/dist-packages/pygal/_compat.py", line 97, in timestamp

return (x - datetime(1970, 1, 1, tzinfo=utc)).total_seconds()

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

```

I propose something like:

```
if isinstance(x, datetime.Datetime):
return (x - datetime(1970, 1, 1, tzinfo=utc)).total_seconds()
else:
return x.value/1e9 #conversion from pandas.Timestamp to EPOCH in seconds
```

This does not lead do an extra dependency for pandas and should solve my issue.

Start in pygal/graph/time.py and pygal/_compat.py, following the datetime_to_timestamp call shown in the traceback. Reproduce the failure with a UTC-localized pandas Timestamp and compare it with a regular datetime, then inspect the existing compatibility tests. Done means DateTimeLine renders both supported timestamp forms without the timezone subtraction error.

Written by the indexing model from the issue text.