Python client: re-connect event-stream, when connection dropped

[masipauskas]

• With grpc stream response, in case of connection drops - instead of getting an error - we stop getting events. This PR sets connection in python client such, that client will transparently attempt to reconnect in case no messages were received in last 15 minutes (timeout is configurable).
• Also now now queue and job_set_id - are both required parameters for cancel / reprioritize requests (in line to actual go API).

In future, we will:

• Intent is to add heartbeat messages from Armada API at a later stage, which would allow us to use a tighter timeout in the future (currently, we indirectly depend on Utilisation event messages, sent every 5 minutes as a proxy for connection heartbeat).
• Reduce redis dependency (for non-event stream use-cases)