Send CRITICAL state to NSCA on backup failure

[timvaillancourt]

If the tool hits the exception handler in Main.py we do not send a 'failed' state in the Notify NSCA method(!). Instead we are sending a success (Nagios state: OK).

1. Start 'Notifier' component at init-time instead of the starting of the backup so we can react to any failure properly. Moved to it's own method .setup_notifier() in Main.py.
2. Fixed incorrect fields passed into Notify on failure in Main.py .cleanup_on_failure().
3. Added more detail to failure message string.
4. Catch exceptions properly in Notify/Notify.py.
5. Remove 'timer' start/stop and 'completed' boolean in Notify/Notify.py - it's already handled by 'Stage'.