Separate Cluster start-up and shutdown from Job submission

[nickbattam-tessella]

The MPI framework job submission process includes calls to 'start up' and 'shutdown' the target cluster.

The start-up and shutdown calls must be moved out of the individual job submission code to remove the significant time overhead this introduces if multiple jobs are submitted, and enable the framework to work with a persistent cluster.

Outcome:

• [ ] Individual jobs do not shut down the cluster after completion
• [ ] Individual jobs do not start up the cluster before starting
• [ ] Jobs will raise an error if the target cluster is not running when the job is submitted

[abuts]

3 is currently implemented. Job starts only if cluster reported ready (as part of cluster start_up process)