Real-time instance monitoring and notification-based alarm management

OpenStack Ceilometer already provides some functionality to monitor and alert the user about faults in the server. It will be useful to enhance this functionality as follows:

[1] Instance State Notification
    This BP proposes adding new compute notification definition regarding instance state to handle event of instance (server) from nova.
    It also enables to create a new meter "instance.state".
    To notify an instance.state change immediately, the BP creates "instance.state" by using the notification agent rather than the pollster.

[2] Event Publisher for Alarm
    The proposal is to create a new event publisher which can send messages to a new alarm evaluator [3].
    The publisher enables the admin to provide event driven notifications to users such that they can achieve fast auto-healing by using this immediate notification mechanism and orchestrator with auto scaling rules.
    Besides the existing Ceilometer usage for billing purposes, this BP enhances Ceilometer to provide additional notification capabilities to the user.

[3] Notification-driven alarm evaluator
    This BP proposes a notification-driven alarm evaluator that is using event notifications received from [2].
    The alarm evaluator does not execute any periodical task, but is triggered by alarm notifications.
    The alarm evaluator will aggregate and correlate different alarms, which will then be notified to the user in order to trigger recovery action(s) on the user-side (e.g. migrate, terminate,re-instantiate etc.).