Fix cgroup handling for systemd with cgroup v2

This fixes issues where kubelet enforces qos and nodeAllocatable on the
worng hierarchy. Kublet will now create the files

/sys/fs/cgroup/kubepods/{burstable,besteffort,}/pod-xyz

when running with systemd as the driver, making it impossible to enforce
the limits on nodeAllocatable.

pkg/kubelet/cm/cgroup_manager_linux.go

@@ -577,7 +577,7 @@ func (m *cgroupManagerImpl) Update(cgroupConfig *CgroupConfig) error {
 
 	unified := libcontainercgroups.IsCgroup2UnifiedMode()
 	if unified {
-		libcontainerCgroupConfig.Path = cgroupConfig.Name.ToCgroupfs()
+		libcontainerCgroupConfig.Path = m.Name(cgroupConfig.Name)
 	} else {
 		libcontainerCgroupConfig.Paths = m.buildCgroupPaths(cgroupConfig.Name)
 	}