Introduce exponential backoff strategy for starting Connection to RabbitMQ

[jollopre]

The existing approach to retry initial connection pauses 1 second before retrying. In the event of RabbitMQ being already distress, simultaneous retries from several processes might contribute to further degradation. In order to prevent that to happen, we'd like to introduce exponential backoff strategy so that clients wait progressively longer between retries.

The solution implemented has to consider the following:

• Using a base for which retry intervals do not grow considerably as the number of retries increases. For instance, choosing a base below 2
• Incorporating jitter, so that avoids collision retrying a connection through simultaneous clients
• This resilience pattern should be only applied when harmoniser is the server, i.e. is the main process running
• This pattern has to be removed for declaring topology and publisher connection.

The existing implementation is implemented here