Add degradation tests into CI

[AKorpusenko]

Added a new script for CI which:

• Runs go test -bench per each of files listed in the config.yml

• Per each of the Benchmark test it's optional to specify allocations count and operations count delta

• Script requires yq and benchstat tools for work

• Each test case is executed 10 times. This is required minimum to get at least somewhat accurate values.

Makefile

Added two new commands:

• degradation-test -- runs the tests with benchmarks and compares the results with the existing in the repo
• update-benchmarks -- a manual Github Actions workflow which can we triggered to update the benchmarks in the repo. Should be used for approved optimizations and performance downgrades

[nkryuchkov]

Great job! Just minor comments

[nkryuchkov]

Can we use the testify library instead?

[MatusKysel]

Suggested change

	opThresholds[testCase.Name] = testCase.OpDelta
	allocThresholds[testCase.Name] = testCase.OpDelta

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 49.1%. Comparing base (4966dce) to head (14b6064).
Report is 4 commits behind head on stage.

Additional details and impacted files

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

[y0sher]

This test/benchmark needs to be updated after #1274 and #1358

[AKorpusenko]

Updated

[y0sher]

@AKorpusenko I don't understand how you updated the test but the benchmark results didn't change? especially how this passes on CI since on linux we now use openssl which should be 10x faster than stdlib rsa (which you run locally on your mac).

[AKorpusenko]

@y0sher, we have to compare the benchmarks only between different GHA workflow runs, right? I've added a manual-update-benchmarks GHA workflow for this, but it only can be used after this code will be in master. Before this, I used my own fork of ssv to get linux runs benchmarks.
In the last commit, I've just made you can see significant performance increase. So it works like we expect.

Another thing I've noticed, is that we have queue_test.go which uses b.Log and this logs appear in the benchmarks, what is wrong from my perspective and should be fixed in another PR.