Make WHS more fault tolerant in case when RabbitMQ is not available

[boriska70]

Use case: if RabbitMQ is not available, we return error 500 to client. Most of clients do not make retries and data is just lost. Likely better solution is:

1. Retry publishing with (increasing) timer 2-4 times in case that Rabbit is just restarting
2. If failed, save the data to handle it later. The only place we have for this is etcd (unless we want to create a volume for WHS). Then (periodically?) WHS checks the etcd content and pablish data, when RabbitMQ connection is OK again.
   NOTE: multiple WHS instances insert the same data from etcd? Can be handled in ES but if possible to finde simple solution in WHS - do this.