feat: tune resource requests/limits for e2-small node (2vCPU/2GB)

Reduce application pod footprint and cap monitoring stack to fit
within ~1.7GB allocatable RAM on a single e2-small node.

backend:  100m/256m CPU, 128Mi/256Mi → 50m/200m CPU, 96Mi/192Mi
frontend: 50m/100m CPU, 64Mi/128Mi  → 20m/80m CPU, 48Mi/96Mi
prometheus: add 50m/300m CPU, 200Mi/400Mi limits
grafana:    add 20m/150m CPU, 80Mi/160Mi limits

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>