EPP deployment should specify resource requirements

[liu-cong]

What would you like to be added:

The example deployment https://github.com/kubernetes-sigs/gateway-api-inference-extension/blob/main/config/manifests/inferencepool-resources.yaml should specify resource requests and limits, and provide a guidance on how to configure resource requirements correctly. This will involve some experimenting/benchmarking, and give some guidance on the resource vs. load.

Note the ongoing development of queuing and prefix cache will significantly change the memory footprint.

Why is this needed:

[kfswain]

Mem footprint will definitely change, and be incredibly biased to the features enabled or added to the EPP. Instead of setting resource limits, we should include a guide as to how we came to those values