Fix slurm configuration in prod2313

[sjpb]

• Using Enable use of custom Slurm builds ansible-role-openhpc#163 means openhpc_slurmd_spool_dir has to be specified instead of openhpc_config_extra.SlurmdSpoolDir, so that the role can actually create the spool dir.
• The openhpc_config_extra dict is defined both environments/nrel/inventory/group_vars/openhpc/overrides and in environments/{prod,vtest}/inventory/group_vars/all/openhpc-generic-slurm.yml. So in both environments the desired nrel config is not actually getting applied.
• environments/nrel/inventory/group_vars/openhpc/overrides.yml defines openhpc_config_extra.StateSaveLocation: /var/spool/slurm/slurmctld. Looking at the terraform for prod and vtest neither define volumes. And there's no state share in environments/nrel/inventory/group_vars/os_manila/overrides.yml. Is state on a persistent disk at all in the cluster? If not, we should fix this. appliances_state_dir should be set, then the defaults will do the right thing (once that override is removed). NB: should retrieve the slurmctld state from the current directory BEFORE reimaging the cluster!
• openhpc_packages_extra_nrel -> openhpc_packages_extra which won't be applied when using generic slurm. Also this contains a lot of openhpc-specific packages. Also needs to be combined with the example openhpc_generic_packages provided.
• openhpc_*_dir is defined differently in nrel and vtest - is this required?
• [ ]

[sjpb]

Replaced by PR on correctly-named branch: #18