telemetry configuration vs docs

On Discuss, https://discuss.hashicorp.com/t/nomad-telemetry/8295 we had a report of some unexpected behavior of the telemetry configuration.

The documentation claims that telemetry is disabled by default, but it looks like a set of "core" metrics are published regardless (much more on the server side than the client). Likewise, the `prometheus_metrics` configuration only toggles on Nomad's core, alloc, and/or node metrics, but always publishes the default golang library metrics.

Some quick testing shows the combinations works out like this:

| `publish_allocation_metrics` | `publish_node_metrics` | `prometheus_metrics` | /v1/metrics | /v1/metrics?format=prometheus |
---|---|---|---|---
| false | false | false | core metrics | prometheus library default metrics
| true | false | false | core + alloc metrics | prometheus library default metrics
| false | true | false | core + node metrics | prometheus library default metrics
| false | false | true | core metrics | prometheus library default metrics + server metrics
| true | false | true | core + alloc metrics | prometheus library default metrics + alloc metrics
| false | true | true | core + node metrics | prometheus library default metrics + node metrics

None of the behavior here seems wrong _per se_, but it doesn't match the documentation which says that not having `true` for these values should be disabling telemetry entirely.

Start with the telemetry configuration documentation and the /v1/metrics and /v1/metrics?format=prometheus endpoints described in the issue. Compare the documented defaults with the behavior in the table, then update the documentation or clarify the configuration semantics so the two agree across the listed combinations.

Written by the indexing model from the issue text.