Improve consistency of labels/tags on Nomad telemetry

### Nomad version
0.12.1

### Operating system and Environment details

### Issue
I think the use of labels/tags should be made more consistent in Nomad telemetry metrics. In particular, I would like to see datacenter and namespace added to all metrics for which those are meaningful. Currently, I do not think they are. For instance, nomad_client_allocs_running does not have either tag, nomad_client_allocations_running has datacenter, but not namespace, and nomad_client_allocs_memory_usage has namespace, but not datacenter. But I believe that all of these metrics are coming from Nomad clients and should therefore all have datacenter. It is possible that some could not have namespace if they are not connected to a specific namespace.

But we should strive to include any relevant or applicable labels/tags on all metrics. Doing this will increase the value of the telemetry metrics by making it easier for customers to build custom dashboards (e.g., in Grafana) that filter metrics by namespace or datacenter.

### Reproduction steps
Run the AutoScaling remote demo: https://github.com/hashicorp/nomad-autoscaler/tree/master/demo/remote
Add the attached dashboards.

Start with the AutoScaling remote demo and attached dashboards, then inspect the telemetry metrics named in the issue: nomad_client_allocs_running, nomad_client_allocations_running, and nomad_client_allocs_memory_usage. Done means relevant metrics consistently expose datacenter and namespace labels where applicable, including handling metrics without a specific namespace.

Written by the indexing model from the issue text.