Issue 5337: Handle closed event loops

[Onkelborg]

Motivation:

The issue that this attempts to correct is #5337

There is a race condition in case someone attempts to send a message on the eventbus to an address at the same time as the receiver is being undeployed - if the receiver's context has it's own worker pool, that worker pool will get shutdown, and it won't be possible to enqueue the operation to receive the message, and exception will be thrown. This affects both send and publish - in the latter case, an undefined number of registered consumers will receive the message in case this race condition affects one of the receivers.

I've written unit tests for both local and clustered eventbus, and I think that the code coverage for them are quite comprehensive.

Conformance:

• Eclipse Contributor Agreement has been signed.
• I think I've managed to following the code style guidelines