[Discovery] Ensure event bus events can't be lost when producing

[robrap]

We may end up with multiple types of solutions that are applicable in different solutions, but we should know more about how to ensure events can't be lost when producing.

Questions:

• What are the various ways that Confluent recommends?
• Is the Outbox pattern a viable solution, and what would it take to implement, including parallel-processing of the outbox itself.
• Is producing using sync instead of async acceptable in any situations as a potentially easy way?
• Do different ack configs help with this, or does this just handle a certain class of problems?
• What does happen now when a server is stopped? Does the batch make it to Confluent?
  • Relates to: [Producer] Make sure we don't lose events on producer shutdown #11

[robrap]

This ticket may have been displaced by openedx/openedx-events#251. Once movement is made on that, it may make sense to close out this ticket.