Change liveness probe so it doesn't fail if there are problems with alertmanager

[edda]

Currently the supernova pods' liveness probe fails if Alertmanager isn't reachable or returns a 50x error. Supernova should behave in such a way that if there are problems with Alertmanager Supernova itself should keep running and just show some sort of message that alertmanager isn't available.