fix: keep the NATS task subscriber consuming after errors

[wolf4ood]

What this PR changes/adds

NatsSubscriber:

• Fetch errors. An exception while fetching messages (e.g. a NATS disconnect) is logged and the subscriber tries
  again after a pause (1 s by default). Before, it ended the consumer loop silently: the subscriber stopped consuming
  for good, with nothing in the logs.
• Ack/nak errors. An exception from nak() while handling a failed message is logged, and the other messages
  of the batch are still handled.
• Stopping. Errors after stop() (e.g. from the closed connection) end the loop quietly.
• Unrecoverable errors (Error) are logged before the loop ends.
• isHealthy(). The subscriber is healthy until it is started; after that, while its loop is running and its
  connection is not closed for good (jnats gave up reconnecting).

NatsTransferProcessTaskSubscriberExtension and NatsContractNegotiationSubscriberExtension register a liveness check
that reports the subscriber as not live when it is no longer consuming messages, so that the runtime can be
restarted.

Why it does that

A subscriber whose loop had stopped left every new task unprocessed, while the runtime looked healthy.

Further notes

• New NatsSubscriberTest. Without the fix, the tests for a failing fetch, a failing nak() and an Error in the
  handler fail.
• The extension tests check that the liveness providers are registered.
• How handler failures map to ack/nak/term is unchanged. Retrying them with a delay is the next item of the issue.

Who will sponsor this feature?

Please @-mention the committer that will sponsor your feature.

Linked Issue(s)

Part of #6116

Please be sure to take a look at the contributing guidelines and our etiquette for pull requests.