Support cgroups v2

The code has been refactored with a slightly simpler architecture,
where the metrics collection frequency is now controlled directly
by Prometheus scrapes. Only the list of cgroups is updated on a
slower interval (as it requires a filesystem scan of /sys/fs/cgroup).

New code has been added to support the cgroups v2 filesystem hierarchy.
2 jobs for v2 in 51 seconds (queued for 6 seconds)