Nova has a `ZoneScheduler` that can run instances on a particular set of
hardware. This attribute hasn't been documented in the api spec, but is
effective at setting the availability_zone and letting ZoneScheduler work.
This fixes how paths and regexps work for less common cases (eg: having
and auth endpoint that is /v2.0/tokens.json).
Serialization to/from json and setting content-type header are added.
No longer silently converts nova api endpoints to 1.1, but errors instead. You
should be using keystone's service catalog in conjunction with
@openstack_compute_service_name.
