Right-size CPU requests cluster-wide and remove missed CPU limits

Increase requests for under-requested pods (dashy 50m→250m, frigate 500m→1500m,
clickhouse 100m→500m, otp 100m→300m, linkwarden 25m→50m, authentik worker 50m→100m).

Reduce requests for over-requested pods (crowdsec agent/lapi 500m→25m each,
prometheus 200m→100m, dbaas mysql 1800m→100m, pg-cluster 250m→50m,
shlink-web 250m→10m, gpu-pod-exporter 50m→10m, stirling-pdf 100m→25m,
technitium 100m→25m, celery 50m→15m). Reduce crowdsec quota from 8→1 CPU.

Remove missed CPU limits in prometheus (cpu: "2") and dbaas (cpu: "3600m") tpl files.

stacks/platform/modules/nvidia/main.tf

@@ -613,7 +613,7 @@ resource "kubernetes_daemonset" "gpu_pod_exporter" {
 
           resources {
             requests = {
-              cpu    = "50m"
+              cpu    = "10m"
               memory = "128Mi"
             }
             limits = {