Allow datacenter and node class filtering

[lgfa29]

The initial implementation for the node filtering would only take either node class or datacenter as the filter input. But as explained in #531, there are valid use cases where defining both is required in order to properly match Nomad clients with cloud instances.

This PR creates a new ClusterNodePoolIdentifier called combinedClusterPoolIdentifier that can be used to merge other identifiers, applying a simple and or or logic.

When both node_class and datacenter are present in a scaling policy, the Autoscaler will use this combined identifier to merge the existing NewNodeDatacenterPoolIdentifier and NewNodeDatacenterPoolIdentifier logic with the and operator.

In the future we may add an option to use the or operator, but it's not clear to me if there is any real benefit since it would increase the node pool instead of reducing it. Regardless, the capability is there if needed.

Closes #531

[jrasell]

Great addition!

Minor comment inline but certainly not a blocker and I expect the answer is future proofing.