Astarte_E2E: handle alerting when rabbit crashes

[matt-mazzucato]

Currently the whole end-to-end testing procedure of Astarte_E2E relies on 1) a connected client, 2) a connected device.
If rabbitmq crashes the device connection is lost and, being the task blocking (see the related astarte-device-sdk-elixir issue), the end-to-end testing gets stuck.

Once the device connection request will be fixed as to be non-blocking, crashes of rabbitmq will be responsible for the reception of the following message {:DOWN, ref, :process, :pid, :normal}, and a consequent inability of assessing Astarte's health.
This behavior must be further investigated.