ref(api): Use requests == limits for better QoS class

[iambriccardo]

Summary

• Default replicator CPU and memory limits to the final request values to help pods qualify for Kubernetes Guaranteed QoS.
• Keep explicit pipeline-level replicator limits as the opt-out path for workloads that need extra headroom.
• Set Vector sidecar limits equal to requests too, since every container in the pod must satisfy request == limit for Guaranteed QoS.
• Update comments, config docs, unit expectations, and k8s snapshots.

[coveralls]

coverage: 72.901% (-0.07%) from 72.972% — add-same-limits into main