Prevent panic in the message handler from terminating the entire service

This problem is present in every microservice that consumes messages from RabbitMQ. The problem has been mitigated in the newest version of Alice, so no service will crash entirely anymore, but there is no indication for the user that something went wrong, so we should change that.

The problem boils down to the fact that a panic (nil pointer dereference, out of range array access etc) bubbles up the call stack until it reaches the main function, where it terminates the entire service. To fix the problem we can simply intercept the panic, and send a message to the frontend saying something went very wrong.