Initiate safemode when connection to k8s fails

[Expected]
At the moment, when k8s is not reachable the hyperflow server enters into safe mode allowing only readonly operations. The same mechanism should happen when k8s is unreachable

[Actual]
When k8s watcher fails, it is apparent that k8s is unreachable but still the server accepts new pod requests. 