Add handling of Kubernetes cluster disconnection

Current state:

• After only one cluster has been registered, it cannot be accessed anymore due to network issues (can be mimicked by removing the floating IP of a Minikube VM on OpenStack).
• If the user C/U/D an application, the Application Controller will raise on error, and the Application will have the reason "Internal error", the same for every general exception.

Proposed feature:

• if the connection is not possible, a specific exception (probably aiohttp.client_exceptions.ClientConnectorError in this case) could be caught and handled
• handling could include a change of the Cluster status (or state). When an Application is C/U/D on the cluster, because of the change of state, a test of connectivity could be performed.

Additional Note

This should probably be handled by a Kubernetes cluster observer, see also comments in #296 (closed).