CleanupBindings before leaving gossip cluster

Turned out that leaving the gossip cluster before cleaning up the bindings
does not propagate the appropriate table events leaving other nodes with stale
entries.
This commit anticipate the logic to cleanup the bindings before leaving the
gossip cluster

Signed-off-by: Flavio Crisciani <flavio.crisciani@docker.com>

network.go

@@ -985,12 +985,13 @@ func (n *network) delete(force bool) error {
 	// before deleting the network from the store since service
 	// bindings cleanup requires the network in the store.
 	n.cancelDriverWatches()
+
+	c.cleanupServiceBindings(n.ID())
+
 	if err = n.leaveCluster(); err != nil {
 		logrus.Errorf("Failed leaving network %s from the agent cluster: %v", n.Name(), err)
 	}
 
-	c.cleanupServiceBindings(n.ID())
-
 removeFromStore:
 	// deleteFromStore performs an atomic delete operation and the
 	// network.epCnt will help prevent any possible