Context

We've had issues with changes creating unintended performance regressions, either in CPU time or memory usage or both. We have benchmarks that can detect performance regressions in many of the cases, but we don't run them regularly or automatically. We should create something similar to the "API Change Report" that runs some or all of the Go driver microbenchmarks and posts a comment with the before/after diff of the PR.

Definition of done

• Create an automation that runs some or all of the Go driver microbenchmarks before and after the PR change, then uses benchstat to compare the results, and posts the results as a comment on the PR.

Pitfalls

• Could make the PR build take longer, depending on how many benchmarks we decide to run.