feat: add Prometheus metrics with OTel sidecar

[robertodauria]

• Add metrics package exposing token request counter on :9990
• Add OTel Collector sidecar to scrape and push metrics to Prometheus
• Update cloudbuild.yaml to build/push OTel image and deploy via service.yaml
• Parallelize Docker builds for faster CI

[coveralls]

Pull Request Test Coverage Report for Build 21258316463

Warning: This coverage report may be inaccurate.

This pull request's base commit is no longer the HEAD commit of its target branch. This means it includes changes from outside the original pull request, including, potentially, unrelated coverage changes.

• For more information on this, see Tracking coverage changes with pull request builds.
• To avoid this issue with future PRs, see these Recommended CI Configurations.
• For a quick fix, rebase this PR at GitHub. Your next report should be accurate.

Details

• 0 of 20 (0.0%) changed or added relevant lines in 2 files are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 0.0%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
handler/handler.go	0	8	0.0%
main.go	0	12	0.0%

Totals
Change from base Build 21243617651:	0.0%
Covered Lines:	0
Relevant Lines:	123

---

💛 - Coveralls

[robertodauria]

@nkinkade FYI

[robertodauria]

Load testing highlighted an issue: multiple instances of the same revision on Cloud Run had exactly the same labels and were overwriting each other when exported by otel-collector. I slightly changed the config by adding GCP resource detection, which introduces several labels to Prometheus metrics including the unique instance ID of each instance. This appears to have solved the problem - e.g., there are no unaccounted for requests when querying Prometheus.

[bassosimone]

Load testing highlighted an issue: multiple instances of the same revision on Cloud Run had exactly the same labels and were overwriting each other when exported by otel-collector. I slightly changed the config by adding GCP resource detection, which introduces several labels to Prometheus metrics including the unique instance ID of each instance. This appears to have solved the problem - e.g., there are no unaccounted for requests when querying Prometheus.

And you did this here, right? 611d2bb