Description of problem:

The tekton_pipelines_controller_taskruns_pod_latency_milliseconds metric - as configured with TektonConfig defaults - returns a Prometheus "instant vector". This significantly limits the usability of this metric - Prometheus can only aggregate with `sum`, `max`, `min`, and other aggregators that produce scalar outcomes.

Workaround: None

Prerequisites (if any, like setup, operators/versions):

• OpenShift Pipelines 5.0.5-492 on OCP 4.15
• Deploy with Prometheus monitoring enabled.

Steps to Reproduce

1. Deploy OpenShift Pipelines v5.0.5-492, with Grafana
2. Set up a Grafana or other visualizer to plot an aggregated view of the tekton_pipelines_controller_taskruns_pod_latency_milliseconds

Actual results:

Cannot use `rate` - it appears that the latency milliseconds is not reported as a Prometheus counter.

Expected results:

Pod (schedule) latency is reported with a counter metric (or equivalent from a histogram) so that we can observe rates over time

Reproducibility (Always/Intermittent/Only Once):

Always

Acceptance criteria: 

Definition of Done:

Build Details:

OpenShift Pipelines 5.05-492
OpenShift 4.15.31

Additional info (Such as Logs, Screenshots, etc):