feat: Add cached k8s client

[MikeGoldsmith]

Which problem is this PR solving?

Uses the kubernetes client directly to query resource details (eg pod, service and nodes) is both expensive and inefficient because it does network requests to the cluster each time. This also led to delays in processing events because the cluster API would enforce backoffs to repeated frequent requests.

This PR introduces a cached client using informers which sets up watchers for the resources we care about and creates in-memory caches for us to query instead.

• Closes feat: optimize k8s client #86

Short description of the changes

• Create utils.CachedK8sClient
• Setup and start the cache in main.go
• Replace direct usage of k8s client with new cached version instead
• Re-enable k8s metadata attrs when processing events
• Refactor k8s metadata function to only add metadata for resources it can find (pod, service, etc)

How to verify that this has the expected result

• k8s metadata is now visible on sent events
• There is no slow-down of events because of API query backoffs

[vreynolds]

k8s attributes are happening without increase in event handled latency 🌮