Add support for creating subfolders in cgroupv2

[DanielShaulov]

The first commit adds cgroup v2 support for load_with_relative_paths - since get_cgroups_relative_paths already stores the path for the unified hierarchy in cgroup v2 under the empty "" key - we can look that key up in load_with_relative_paths and prepend it to the passed path for all subsystems and for the "path" field in the Cgroup struct itself.
Some other changes in load_with_relative_paths are:

1. Accept relative_paths by reference so it can be reused without copying
2. Accept an empty path - to create a cgroup that represents the "current" cgroup if used with the relative paths from get_cgroups_relative_paths

The second commit adds the ability to "create" a cgroupv2 hierarchy without enabling subtree_control. This is crucial, because in cgroupv2, due to the rule that processes can only appear in leafs, if you want to create a subdirectory of existing cgroup with a process in it you need the following order of operations:

1. create the subfolder
2. move all processes to the subfolder cgroup.procs
3. enable subtree_control on the parent

So with the changes in this PR - the following code creates a directory "foo" under the existing cgroup, moves all processes from current cgroup into "foo" and then enables subtree_control on the parent.

If this is merged I will open another PR for Builder to expose this from builder as well.

let relative_paths = get_cgroups_relative_paths().unwrap();
let cg_parent =
    Cgroup::load_with_relative_paths(crate::hierarchies::auto(), "", &relative_paths);
let cg =
    Cgroup::load_with_relative_paths(crate::hierarchies::auto(), "foo", &relative_paths);

cg.create_disabled().unwrap();
for pid in cg_parent.procs() {
    cg.add_task_by_tgid(pid).unwrap();
}
cg.create().unwrap();

[DanielShaulov]

@Tim-Zhang any chance for a review?

[fidencio]

@DanielShaulov, please, could you make sure we have the Clippy Check happy?

[DanielShaulov]

@fidencio fixed - can you approve the workflow please?

[Tim-Zhang]

@DanielShaulov sorry, I have't been paying attention to this repo recently. how about resolving the confit first?

[DanielShaulov]

@Tim-Zhang - done, and make test and make clippy both pass for me

[justxuewei]

@DanielShaulov Please rebase your code, we don't allow merge commits.

[DanielShaulov]

@justxuewei done

[Tim-Zhang]

Close for triggering the CI

[Tim-Zhang]

@justxuewei PTAL and there still are some problems with the CI @DanielShaulov

[Tim-Zhang]

How can we assume the key "" always exists? if the user don't get relative_paths by calling get_cgroups_relative_paths or the code run on cgroups v1, it may be panic here.