ceilometer reporting is not functional

[Adam Gandelman]

While converting the messaging layer to oslo.messaging, I noticed the current ceilometer stats reporting is a bit busted.

First: The ReadStats state is responsible for collecting bandwidth stats (via a bandwidth callback on the vm_manager). I dont think anything ever causes transition into this state. This could be triggered via a new event that could be pushed onto the queue either by a RUG periodic task or something external (rug-ctl on a cronjob?)

second: The new VM manager doesn't actually define a bandwidth callback, so if ceilometer is enabled and you somehow trigger this state transition, it doesn't actually do anything (or there may even be an exception, I dont remember of the top of my head)

third: We need to define some message format for broadcasting stats, and define what we want to track. It looks like currently/historically this has been used to report per-router bandwidth stats to ceilometer. It might be good to extend this to report general RUG stats (ie, events per minute), and to standardize on a messaging format.