Unable to handle a CgroupV2 system with a CgroupV1 controller

[DanielleMaywood]

It appears it is possible to have a CgroupV2 filesystem that also uses a CgroupV1 controller.

It has been discovered that a certain version of Docker can result in a /proc/self/cgroup looking like:

1:net_cls:/
0::/

A typical CgroupV2 setup would only contain a single line 0::/path/of/group. We need to be able to handle this use-case.

This does raise a question of: Could a CgroupV2 system have a CgroupV1 CPU and memory controller? If so, we'll have to do another refactor.

The simplest solution is to fix our parsing of /proc/self/cgroup and assume a CgroupV2 CPU and memory controller.