Implement retry policy on another backend in case of failure

[hamadodene]

When a request arrives, the backend is chosen randomly among the backends configured in the action associated with the route.

When the backend is chosen and the backend becomes unresponsive due to problems we immediately return a "Service Unavailable" error.

For the same route there could be multiple backends therefore a retry policy could be implemented so that if a backend does not respond the request is passed on to another backend.

This should help deal with cases where there are communication issues with a particular backend.

It should also be possible to indicate on which method requests to retry as for PUT,POST ect. it may not be appropriate in some cases.
While for GET requests it should always be possible to perform a retry.

For PUT,POST type requests, the type of error should be evaluated. If the error is during the connection phase with the backend it might be possible in this case to perform the retry.

[hamadodene]

Something like nginx https://nginx.org/en/docs/http/ngx_http_upstream_module.html

@pv3nturi @aluccaroni what do you think about?