Modify k8s API caches and watchers to only store relevant projections of pod

[AjayTripathy]

Is your feature request related to a problem? Please describe.
#2637

Describe the solution you'd like
We don't use all the data out of the k8s API for each pod. Cutting out unnecessary fields can dramatically reduce the memory profile of opencost. However, I'm not sure if that's supported out of the box in the k8s caching watchers; more investigation would be needed.

Estimated impact:

Opencost seems to keep almost every Kubernetes object fully in memory, but only a few fields are used. The watcher and cache took up 56% of the memory, while the metrics emitter used only 4%.

I think we could probably halve the cache memory... so 25% improvement