JobManager: add resilience for backend failures

[jdries]

Backends have all kind of intermittent failures that go away over time. This usually means they send 50x errors.
Especially for the job manager, it can be better to keep trying until the backend works again. The python requests library can be configured to do this.
This way, long running tasks are more resilient.

[jdries]

FYI, my current approach to add retries in connection.py:

        from requests.adapters import HTTPAdapter, Retry
        retries = Retry(total=5,read=50,other=50,status=50,
                        backoff_factor=0.1,
                        status_forcelist=[ 502, 503, 504, 404],
                        method_whitelist=["HEAD", "GET", "OPTIONS","POST"])
        self.session.mount('https://', HTTPAdapter(max_retries=retries))
        self.session.mount('http://', HTTPAdapter(max_retries=retries))