Add retries for rmq publish

[rob-looksrare]

Publishing
* graceful restart - ok as is. Commit finishes when last event gets published.
* kill - ok as is. Causes ord to revert to previous savepoint which takes a long time to repair when starting up, but once it's up it re-emits the blocks that were not committed so not event loss.
* rmq failed ack - we retry to create a new connection, if not successful we gracefully shutdown the app and have event loss.
* rmq connection killed - we retry to create a new connection, if not successful we gracefully shutdown the app and have event loss.
* rmq down - we retry to create a new connection, if not successful we gracefully shutdown the app and have event loss.

This should cover most of the cases where we have infra issues.

Reindexing from scratch on the light version takes ~8 hours so we want to avoid it and worst case when rmq goes down we need to spin up a new instance and reindex til the failure.

Consumers

• graceful restart - ok as is, event goes back to the queue.
• kill - ok as is, event goes back to the queue.
• processing error - we requeue 3x and reject the message after.

[rob-looksrare]

Before merging this need top stop mainnet and let the queues drain in order to recreate them with "x-queue-type": "quorum"