Fixes https://github.com/coder/coder/issues/12721
If a container in docker is started with `--cgroupns=private` (which is
the default behaviour in docker) then `/proc/1/cgroup` has the following
content:
```
0::/
```
If a container in docker is started with `--cgroupns=host` then
`/proc/1/cgroup` has the following content (hash will vary):
```
0::/docker/aa86ac98959eeedeae0ecb6e0c9ddd8ae8b97a9d0fdccccf7ea7a474f4e0bb1f
```
Currently we are determining if a host is containerized by assuming the
second scenario. This means the existing behaviour of sniffing
`/proc/1/cgroup` is not always sufficient for checking if a host is
containerized.
According to [the cgroups(7)
man-page](https://man7.org/linux/man-pages/man7/cgroups.7.html) there
exists a `cgroup.type` file in a nonroot cgroup. This exists in Linux
versions after `4.14`.
> Linux 4.14 added thread mode for cgroups v2.
> With the addition of thread mode, each nonroot cgroup now contains a
new file, cgroup.type
This means we can check for the existence of
`/sys/fs/cgroup/cgroup.type` to see if we are in a container or not.
- For cgroups v1 the wrong cgroup file was being read
to determine max memory. This commit updates the file
from '/sys/fs/cgroup/memory/memory.max_usage_in_bytes' to
'/sys/fs/cgroup/memory/memory.limit_in_bytes'
* Attempts reading cgroupv1 quota, period, usage from /sys/fs/cgroup/cpu,cpuacct by default
* Fall back to /sys/fs/cgroup/cpu for v1 quota and period
* Fall back to /sys/fs/cgroup/cpuacct for v1 usage
Fixes https://github.com/coder/coder/issues/8468
Danielle Maywood