Don't orchestrate start if the instance is local_expect=="started"

So an instance evaluating as "stdby up" when a monitored resource goes down
(the only resource determining the service is up instead of stdby up), the
orchestrator does try an instance start, and the resource orchestrator can
enter the resource restart and TOC codepath.

lib/osvcd_mon.py

@@ -500,6 +500,8 @@ def service_orchestrator_auto(self, svc, smon, status):
         """
         if not self.compat:
             return
+        if smon.local_expect != "started":
+            return
         if svc.frozen() or self.freezer.node_frozen():
             #self.log.info("service %s orchestrator out (frozen)", svc.svcname)
             return