Fix behavior of cgroups path discovery

[fearful-symmetry]

What does this PR do?

Fixes #132
and fixes #139

This sanitizes the "relative" namespace mounts that we get when we monitor the host system from within newer versions of docker.

This also adds much more complex logic for setting the v2 rootpath we use for fetching metrics from v2 cgroups paths.

This also aggressively cleans up the unit tests in cgroup because they were a mess

With this change, cgroups metric collection works properly on docker configs where where the container is using a private namespace. This value is configurable from docker 1.41+

Checklist

• My code follows the style guidelines of this project
• I have commented my code, particularly in hard-to-understand areas
• I have added tests that prove my fix is effective or that my feature works
• I have added an entry in CHANGELOG.md

How to test this PR

• run go test ./tests from the repo root
• build metricbeat with this version of elastic-agent-system-metrics, run system/process metrics.

[fearful-symmetry]

Decided to just fix #139 while I'm at it, so I can at least run the tests.

[fearful-symmetry]

Okay cleaning up the unit tests is harder than I thought

[fearful-symmetry]

Alright, the fix is still kind of broken for cases where we're trying to monitor a process on the host system that doesn't belong to a root cgroup. I think we need to use the host paths to find the actual cgroup on the host system, and then construct the bath inside hostfs from within the container.

[fearful-symmetry]

So, tests are finally passing. One thing I'd like to add is some kind of manual bypass/setting for ContainerizedRootMount, as a sort of failsafe, in case there's some weird edge case I haven't been able to tease out.

[leehinman]

Looks ok.

Can you add some examples to the PR description about what is now supported and any details on which versions of docker this applies to.