heat: Update addresses on CREATE_FAILED

This is required for Tempest CI to fetch master/node addresses in order
to collect logs from them on cluster creation failure.

Change-Id: I24ac7ff632a8758bfefa5b66341a19eb9712dac6
(cherry picked from commit 339a771587)

magnum/drivers/heat/driver.py

@@ -528,6 +528,7 @@ class HeatPoller(object):
                 self.nodegroup.stack_id, resolve_outputs=False)
 
             if stack.stack_status in (fields.ClusterStatus.CREATE_COMPLETE,
+                                      fields.ClusterStatus.CREATE_FAILED,
                                       fields.ClusterStatus.UPDATE_COMPLETE):
                 # Resolve all outputs if the stack is COMPLETE
                 stack = self.openstack_client.heat().stacks.get(