When a port binding resource update comes in from the provisioning
queue, its host will always be the FQDN (if Neutron/Nova have been
configured to use FQDN hostnames.) If use_fqdn is false, we already
use the short host name on create and in tracking existing neutron
and CVX resources, but on delete, we're correctly querying the DB
with the FQDN, but then continuing to use the FQDN in forcing the
CVX update and generating the data for the HTTP DELETE call.
The proposed fix is to override the force_resource_update and
_delete_neutron_resource methods to correct the hostname in those
calls if necessary.
(cherry picked from commit