refactor(nodeadm): rework containerd template mixin interface

[ndbaker1]

Issue #, if available:

Description of changes:

In order to recognize when the running instance has GPU devices, we currently use a hard coded list of instance type families. to keep this list updated or remove it we can do one of the following:

1. use workflows to generate PRs for the changes. this isn't so bad but introduces a build time dependency
2. make api calls to ec2.DescribeInstanceTypes to get whether the current instance has a GPU device
3. read local data on the instance to determine if there is an nvidia device attached

this PR implements (3) from above by looking for the nvidia vendor name in /proc/bus/pcie/device. the original logic is left in and PCIe devices are used as a fallback.

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

Testing Done

See this guide for recommended testing for PRs. Some tests may not apply. Completing tests and providing additional validation steps are not required, but it is recommended and may reduce review time and time to merge.

[cartermckinnon]

Bit too much refactoring mixed in with the logical changes, IMO; I think this PR can be split

[cartermckinnon]

Name is a little clunky, I liked mixin in the prev impl 😉 or maybe modifier?

[ndbaker1]

ill try going with modifier

[cartermckinnon]

What's this about? Doesn't seem related to the other changes

[ndbaker1]

The top level daemon config calls should contain separate components and i didn't like how the runtime spec was nested inside the config.toml call.

i think the kubelet daemon sets the right precedent

amazon-eks-ami/nodeadm/internal/kubelet/daemon.go

Lines 28 to 44 in 28f65e2

	func (k *kubelet) Configure(cfg *api.NodeConfig) error {
	if err := k.writeKubeletConfig(cfg); err != nil {
	return err
	}
	if err := k.writeKubeconfig(cfg); err != nil {
	return err
	}
	if err := k.writeImageCredentialProviderConfig(cfg); err != nil {
	return err
	}
	if err := writeClusterCaCert(cfg.Spec.Cluster.CertificateAuthority); err != nil {
	return err
	}
	if err := k.writeKubeletEnvironment(cfg); err != nil {
	return err
	}
	return nil

[cartermckinnon]

I don't think this should be an error; configurator-s should be composable or there's not much benefit to the abstraction.

[ndbaker1]

I agree this looks funky, mainly was a byproduct of trying to continue using the abstraction without a great reason. It doesn't work like a real mixin, it just populates a set of parameters and wouldn't make sense if they got written more than once

I think ill just call them in a loop if it matches rather than doing this additional aggregation + check

[ndbaker1]

#2146 replaces this PR for PCIe detection functionality, while this will be turned into a refactor for the mixin interface,

[ndbaker1]

superceded by #2227