Drop the use of magic openstack project_id

When we put validation into the project_id in urls we had to expand
the default set of allowed values to include 'openstack' and
'fake'. This cleans up all the instances of 'openstack' in the
functional tests, and api samples, and instead uses a constant string
which looks like a real project_id.

We need a stable project_id and not a sentinel otherwise the samples
will be arbitrarily changes on every update.

Change-Id: I1099fd43d2374abd1658b0917765d5375c4b8d1d

doc/api_samples/os-server-groups/v2.13/server-groups-get-resp.json

@@ -5,7 +5,7 @@
         "policies": ["anti-affinity"],
         "members": [],
         "metadata": {},
-        "project_id": "openstack",
+        "project_id": "6f70656e737461636b20342065766572",
         "user_id": "fake"
     }
 }

doc/api_samples/os-server-groups/v2.13/server-groups-list-resp.json

@@ -6,7 +6,7 @@
             "policies": ["anti-affinity"],
             "members": [],
             "metadata": {},
-            "project_id": "openstack",
+            "project_id": "6f70656e737461636b20342065766572",
             "user_id": "fake"
         }
     ]

doc/api_samples/os-server-groups/v2.13/server-groups-post-resp.json

@@ -5,7 +5,7 @@
         "policies": ["anti-affinity"],
         "members": [],
         "metadata": {},
-        "project_id": "openstack",
+        "project_id": "6f70656e737461636b20342065766572",
         "user_id": "fake"
     }
 }