/sig scheduling
/kind cleanup

scheduler-perf has few tests for itself at the moment; we roughly check the behavior with -tags performance, short or -tags integration-test, short, which could miss some potential issues with it.
scheduler-perf right now serves a crucial role for the scheduler to keep our targets in various metrics, and we're adding more and more features into it. Unless we have proper testing for it, the more features we add into it, the more likely we could break some existing parts without noticing it since the reviewing would get more and more difficult.

So, this issue aims to add tests for it. I guess it also requires some refactoring, especially around runWorkload, which is currently huge.

/cc @alculquicondor @macsko @dom4ha