Refine Engine <-> Executor protocol

From https://etherpad.openstack.org/p/juno-summit-mistral:
* Scalable engine executor interface
        a) change the name to worker, ActionRunner?
        b) worker parallelism (a worker should be able to handle more than one task in parallel)
        c) engine-executor communication, clean-up interface and STATE
        d) parallelism engine-executor
        - Message queue, DB, REST api - convey_task_result()
        e) ActionRunner shouldn't have access to DB (e.g. in a case when AR is behind a firewall)

Please also refer to https://etherpad.openstack.org/p/mistral-executor-protocol to get the details of Kirill's research.