Running benchmarks reliably with GitHub action

This repo shows this workflow:

1. Run benchmarks for your Go code (func Bench...) in the main branch and in PRs.
2. On every commit to a PR, it compares the benchmark results against main:
   1. if no main benchmarks are available, just publish benchmarks in the summary of the PR.
   2. if there main benchmarks, and they got worse in the PR, fail the PR: