Service monitor triggered from two callbacks

[dnephin]

tron.core.service.Service has a repair_callback which calls instances.restore() which calls monitor_task.run().

tron.core.serviceinstance.ServiceInstance has a MonitorTask which calls it's queue() every time it completes.

This results in many warnings 'monitor action already exists' and may lead to other errors. I believe that Server.restore() should only recreate instances, and should let the instance take care of running the monitor on its own.