libct/cgroups/fs: code cleanups

[kolyshkin]

this is separated out from #2438 in order to make review easier

This contains some minor cleanups and optimizations, but mostly just code removal.

TODO: add
* db45ce7
* 44a02d2
once #2496 is merged

[kolyshkin]

For the reference, here's the benchmark used for the last commit

package fs

import (
	"testing"
)

func initPaths() map[string]string {
	paths := make(map[string]string)

	for _, ss := range subsystems {
		paths[ss.Name()] = "/sys/fs/cgroup/" + ss.Name() + "/user.slice/user-1000.slice/session-881.scope/xx88"

	}

	return paths
}

func BenchmarkIterateOverPaths(b *testing.B) {
	paths := initPaths()

	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		for name, path := range paths {
			ss, err := subsystems.Get(name)
			if err == errSubsystemDoesNotExist || path == "" || ss == nil {
				continue
			}
		}
	}
}

func BenchmarkIterateOverSubsystems(b *testing.B) {
	paths := initPaths()

	b.ResetTimer()
	for i := 0; i < b.N; i++ {
		for _, ss := range subsystems {
			path := paths[ss.Name()]
			if path == "" {
				continue
			}
		}
	}
}

[mrunalp]

This could be of use to external clients and maybe later on if/when we factor this out into a cgroups library.

[kolyshkin]

Theoretically, yes.

Practically,

1. I don't see it being used by kubernetes;
2. It uses d.path which is expensive, it requires to parse mountinfo (this is actually the main point of it being removed);
3. Its API is not in line with e.g. Set (which does not use d.path but the path is provided directly).

Also, if needed, it can be trivial to reimplement / re-add, as it doesn't do anything special, so I see no point in keeping it.