datetime: support timezone bindings

[davidhewitt]

This PR implements safe wrappers for the Python 3.7 timezone bindings described in #207.

Rather than try to do anything complicated with a new type representing exactly datetime.timezone, I return &PyTzInfo from the three new functions timezone_utc(), timezone_from_offset() and timezone_from_offset_and_name(). I think this is good enough to be useful; datetime.timezone is a subclass of datetime.tzinfo so doing this is correct.

I also adjusted a few constructors taking tzinfo arguments to take Option<&PyTzInfo> instead of Option<&PyObject>.

Finally, I tweaked the ffi symbol PyDateTime_TimeZone_UTC I added recently in #1573 to resolve to &'static *mut ffi::PyObject, which I think is a more correct representation of the "unsafe ffi" nature of that global.

Closes #207

cc @pganssle in case you might be interested / have an opinion on this.