Retry keystone_wait_for_propagation() on exception (#175)

At the CI jobs keystone_wait_for_propagation() fails when the
api_version attribute hasn't been propagated yet in the identity-service
relations. This patch will retry automatically giving time to keystone
when the system where this is running is not fast enough.

Fixes LP: #1668954

charmhelpers/contrib/openstack/amulet/utils.py

@@ -40,6 +40,7 @@
 import pika
 import swiftclient
 
+from charmhelpers.core.decorators import retry_on_exception
 from charmhelpers.contrib.amulet.utils import (
     AmuletUtils
 )
@@ -423,6 +424,7 @@ def tenant_exists(self, keystone, tenant):
         self.log.debug('Checking if tenant exists ({})...'.format(tenant))
         return tenant in [t.name for t in keystone.tenants.list()]
 
+    @retry_on_exception(num_retries=5, base_delay=1)
     def keystone_wait_for_propagation(self, sentry_relation_pairs,
                                       api_version):
         """Iterate over list of sentry and relation tuples and verify that