metrics: pre-define telemetry datapoints for easier consumption

The Nomad agent does not pre-define datapoints, so each metric only appears in the API (for example when using Prometheus) during a specific window after it has been triggered. If the datapoint is not triggered again within a certain period, the datapoint will be garbage collected and will cease to appear in the API return object. In a similar manner, if a datapoint is never triggered, it will never show up in the API return.

In clusters with high throughput that utilise all features, this isn't a huge problem. In all but the largest clusters, this becomes a problem, and when setting up dashboards on idle clusters, it makes the task quite tricky.

Example: I am settings up Grafana dashboards to monitor certain Nomad subsystem as part of a project. In particular I want to look at the scheduler performance and our documentation indicates the metric `nomad.worker.invoke_scheduler.` is important. I am not currently utilising all the scheduler types, which means I cannot create graphs for all of them unless I trigger samples jobs to trigger the datapoint.

Ideally, Nomad would pre-define its metrics, or at least some, so the problems described above can be lessened.

Start by tracing Nomad's agent metrics API and its Prometheus representation, using nomad.worker.invoke_scheduler. as a concrete example. The issue names no files or tests, so first locate the metrics implementation and existing coverage; done means the requested datapoints appear for idle clusters without first triggering them.

Written by the indexing model from the issue text.