Reduce downtime during platform stack applies

CrowdSec fixes:
- Increase ResourceQuota requests.cpu 1→4 (was at 302%, blocking upgrades)
- Add LAPI startupProbe: 30 attempts × 10s = 5min startup window
  (LAPI pods were failing default startup probe during rolling upgrades)
- Reduce Helm timeout 3600s→900s with wait=true, wait_for_jobs=true

Prometheus startup guard on 8 rate-based alerts:
- PodCrashLooping, ContainerOOMKilled, CoreDNSErrors,
  HighServiceErrorRate, HighService4xxRate, HighServiceLatency,
  SSDHighWriteRate, HDDHighWriteRate
- Suppresses false positives for 15m after Prometheus restart

stacks/platform/modules/crowdsec/main.tf

@@ -107,7 +107,7 @@ resource "helm_release" "crowdsec" {
   chart      = "crowdsec"
 
   values  = [templatefile("${path.module}/values.yaml", { homepage_username = var.homepage_username, homepage_password = var.homepage_password, DB_PASSWORD = var.db_password, ENROLL_KEY = var.enroll_key, SLACK_WEBHOOK_URL = var.slack_webhook_url, mysql_host = var.mysql_host })]
-  timeout       = 600
+  timeout       = 900
   wait          = true
   wait_for_jobs = true
 }

stacks/platform/modules/crowdsec/values.yaml

@@ -56,6 +56,12 @@ lapi:
       memory: 128Mi
     limits:
       memory: 1Gi
+  startupProbe:
+    httpGet:
+      path: /health
+      port: 8080
+    failureThreshold: 30
+    periodSeconds: 10
   priorityClassName: "tier-1-cluster"
   replicas: 3
   topologySpreadConstraints: