Go Benchmarks in Github Actions

[embano1]

While the newly added benchmark GHA running on PRs to detect regressions gives us some good insights, it is currently hard to parse as we cannot directly alert on a PR due to Github's security model for workflows triggered on pull_request.

IMHO we could simplify the whole benchmark workflow and make use of Job Summaries if we can get a nicely rendered markdown table from the benchmark output. I am using tparse to do this for regular go test but haven't looked into it wrt go test -bench.

I am wondering if we could just adopt tparse or something custom we write to get a markdown table for benchmarks which we print on the job summary and have it fail on threshold X. Thoughts? @timbray @yosiat

[embano1]

Update: there's a tracking issue in tparse: mfridman/tparse#67

[timbray]

Micha, I'm going to trust you on this one, I am still way behind on this CI stuff (catching up a bit, but still). I would like us to have all non-failing actions though.

[embano1]

Sure, will keep track of it so at least we simplify the benchmark review experience (using summaries).

[yosiat]

@embano1 Sounds good to me 👍