ci: add continuous benchmark tracking dashboard

[scottgerring]

Summary

• Adds a continuousBenchmark job that runs on every push to main, storing Criterion results via github-action-benchmark and publishing a dashboard to GitHub Pages. This follows the pattern used in otel-collector-contrib and otel-arrow
• Manually add libssl-dev (fix: add openssl-dev to benchmark container env #3401 is needed to fix benchmarks anyway)
• Mentions the benchmark dashboard in CONTRIBUTING.md

Dashboard will be at: https://open-telemetry.github.io/opentelemetry-rust/dev/bench/

Requires enabling GitHub Pages on the gh-pages branch (the action creates the branch automatically on first run). We'll need to chat with someone in opentelemetry-community to validate this is well and good and do the repo modifications. I will chase this up once we are happy before we merge.

Testing

• Sample dashboard (on fork): https://scottgerring.github.io/opentelemetry-rust/dev/bench/ - FYI I've only run this multiple times for a small subset of benchmarks - the rest need another run to show deltas
• Sample PR showing performance label still triggers comparison job:
  test: verify PR benchmark workflow scottgerring/opentelemetry-rust#1

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 83.2%. Comparing base (a05dbe6) to head (7782280).
⚠️ Report is 9 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #3404     +/-   ##
=======================================
+ Coverage   82.7%   83.2%   +0.4%     
=======================================
  Files        128     128             
  Lines      24811   24899     +88     
=======================================
+ Hits       20526   20716    +190     
+ Misses      4285    4183    -102     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[scottgerring]

Adding performance tag to demonstrate that the regular perf bench still runs.

[cijothomas]

Thanks. (I setup the dashboards for OTel-Arrow repo, so I an figure out the process here.. Its relatively easy, just need to dig up the steps)

Before we do that, I'd like to double confirm that we want to run all the benches on every commit to main. The perf runner is a scarce commodity (there is just 1 machine), and I don't see much value in running every benchmark on every main commit. We need to have a curated subset.

Alternatively, we can just have a nightly or even weekly (or twice a week) run for all benchmarks.

[scottgerring]

Thanks. (I setup the dashboards for OTel-Arrow repo, so I an figure out the process here.. Its relatively easy, just need to dig up the steps)

Before we do that, I'd like to double confirm that we want to run all the benches on every commit to main. The perf runner is a scarce commodity (there is just 1 machine), and I don't see much value in running every benchmark on every main commit. We need to have a curated subset.

Alternatively, we can just have a nightly or even weekly (or twice a week) run for all benchmarks.

@cijothomas ack concern on resource consumption. I think we should err on the side of running everything (else why have the benchmarks) at a lower cadence. It takes about 30 minutes to run presently. I think nightly would be fine but we can tone it down to weekly if you prefer

[cijothomas]

Thanks. (I setup the dashboards for OTel-Arrow repo, so I an figure out the process here.. Its relatively easy, just need to dig up the steps)
Before we do that, I'd like to double confirm that we want to run all the benches on every commit to main. The perf runner is a scarce commodity (there is just 1 machine), and I don't see much value in running every benchmark on every main commit. We need to have a curated subset.
Alternatively, we can just have a nightly or even weekly (or twice a week) run for all benchmarks.

@cijothomas ack concern on resource consumption. I think we should err on the side of running everything (else why have the benchmarks) at a lower cadence. It takes about 30 minutes to run presently. I think nightly would be fine but we can tone it down to weekly if you prefer

Okay. Let us run all of them, and modify the schedule to be nightly, not every commit.
(We can adjust this in the future as we evolve)

[scottgerring]

@cijothomas :

• Updated schedule to fire once daily
• Added a guard to not run if the commit being benchmarked has already been benchmarked; validated here

I reckon this should be good to go; we'll just need to enable github pages on this repo.