Tune benchmark params to reduce variance

[jack-berg]

Despite bare metal runners, the variance for the benchmarks we publish is unacceptably high, with CV's regularly well above 10% and sometimes exceeding 20%: https://open-telemetry.github.io/opentelemetry-java/benchmarks/

I've been doing some experimenting and unfortunately, it seems like in order to get the variance down, we need to increase the JMH forks, and iteration count. I tried to minimize the run time while keeping CV <5%, and landed on:

• Forks: increase from 1 to 3
• Warmup iterations: reduce from 5 to 3
• Measurement iterations: increase from 5 to 10

Runtime change:

• Previous: 1 fork x 5 warmup x 5 measurement x 1s = 10s per case
• New: 3 forks x 3 warmup x 10 measurements x1s = 39s per case
• Cases: 40 metric + 6 log + 6 span = 52
• Expected run time change = 8.6min to 33.8min (in practice, the runtimes are longer)

@trask I know we run on shared hardware. Wondering if we can afford to reduce the variance without impacting other SIGs. If not, we could make up for the increased time by switching from running on every commit to main to running daily.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 90.25%. Comparing base (6336abd) to head (2d48b13).

Additional details and impacted files

@@            Coverage Diff            @@
##               main    #8308   +/-   ##
=========================================
  Coverage     90.25%   90.25%           
  Complexity     7687     7687           
=========================================
  Files           850      850           
  Lines         23198    23198           
  Branches       2354     2354           
=========================================
  Hits          20938    20938           
  Misses         1532     1532           
  Partials        728      728           

☔ 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.