Make KE Runtimes better deal with Knowledge Directory being unavailable.

[bnouwt]

When stopping the knowledge-directory while running the unreachable-runtimes example, the KER gives a lot of errors and seems to not update its OtherKnowledgeBaseStore. Therefore it keeps trying to send messages to KBs that it no longer knows. What behavior would we expect here? Do we want to KERs to keep exchanging messages while the knowledge-directory is unavailable? Because they can easily do that. Or do we want the KERs to remove all RemoteKerConnection's and no longer be able to send any messages to any remote KER?

Also, when the knowledge-directory becomes available again, it does not recover properly, but it remains in the state of not being able to send messages to KBs in other KERs.

[bnouwt]

Eventually the unreachable-runtimes example seems to recover fully, but it seems to take a while.

[bnouwt]

Maybe introduce the following additional Inter-KER REST operations:

• remove all RemoteKerConnection's from the KER and empty the OtherKnowledgeBase. This way, we can temporarily shut down the knowledge-directory and then remove RemoteKerConnection's.
• add mechanism to manage a list of KERs to ignore by a particular KER. But then this ignored KER should also ignore all other KERs, otherwise it will keep sending messages to them.