Fix GPU Integ Tests and add PV, PVC Integ Tests

[sky333999]

Description of changes

1) Make GPU addon test consistent with GPU daemon test

• The GPU addon test previously followed a diff pattern that does some setup via terraform and some setup for patching the agent image in the GH workflow. This PR standardizes the setup by making moving all the setup to terraform.

2) Use actual GPU nodes for GPU daemon test and spin up real gpu-burn

• This allows generating all GPU metrics that rely on a real gpu pod existing (& mapped to a service). The pod is spun up via a statefulset so we can deterministically set the name to match the mock dcgm-exporter data.

3) Update PV & PVC tests

• Add validations for remaining PV & PVC metrics
• Create PV & PVC as part of tf setup so we get real metrics

4) Remove unnecessary UseE2EMetrics flag and clean up related code

License

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

Tests

https://github.com/aws/amazon-cloudwatch-agent/actions/runs/20257453246 shows passing integ tests for GPU/Daemon and GPU/Addon.

[the-mann]

overall really good change, thanks for getting this done. I have a couple suggestions on how we can improve the debugging experience if the test fails again

[the-mann]

a bit nitpicky, but can we use something like

cd "$(git rev-parse --show-toplevel)"

instead?

[sky333999]

Thinking if we should avoid adding a dependency on git for this..
For our GH runners specifically, it should work - but assuming someone has this code available somehow without git, may end up causing some pain.

[the-mann]

i think that's a pretty extreme edge case, no?