benchmarking: expose raw atelet and ate metrics via Prometheus export…

[Nishanth Kotla (Nishanth29)]

Fixes #1589

What this PR does

This PR routes incoming raw OTLP metrics from atelet and ateapi to the Prometheus pull exporter on port 8889 in benchmarking/telemetry/meter.yaml.

Problem

Currently, telemetry-meter only routes aggregated count metrics (substrate_spans and substrate_datapoints) to the Prometheus exporter on port 8889. Raw OTLP metrics (including snapshot sizes, restore durations, and worker pool states) are only forwarded upstream to managed cloud collectors.

In local or standalone benchmarking environments where cloud monitoring backends are not configured, in-cluster Prometheus cannot scrape these raw histograms and gauges.

Proposed Changes

In benchmarking/telemetry/meter.yaml:

• Add a metrics/raw pipeline under service.pipelines routing incoming otlp metrics through the memory_limiter processor to the prometheus exporter on port 8889.
• Leaves existing forwarding and counting pipelines intact.

How this was tested

• Deployed on a benchmark cluster and ran the glutton user workload.

• Verified that telemetry-meter:8889 exposes atelet_snapshot_size_bytes and ate_workerpool_workers.

• Verified that in-cluster Prometheus scrapes these metrics via the default pod annotations without error.

• Tests pass

• Appropriate changes to documentation are included in the PR

[Max Smythe (maxsmythe)]

LGTM