catch performance regressions between versions

[aryszka]

@jonathanbeber showed me this project: https://github.com/knqyf263/cob

We could potentially use it, but of course other solutions, too, to catch small performance regressions between old and new versions of Skipper. We could refactor and extend the scripts in the skptesting package that are currently not part of the regular build, and use go test -bench instead of bash.

[aryszka]

we can use the test dataclient and the root package of Skipper to get the highest possible level of a Skipper process programmatically in the go test specs, and maintain realistic test scenarios: testdataclient.