fix: corrigir tipos de servidor AMD (CPX→CX), autoscaler, stdin e cleanup

- Corrigir tipos Hetzner: CPX não existe, usar CX (Intel shared vCPU)
  - CX23 (2 vCPU, 4GB) equivalente ao CAX11
  - CX33 (4 vCPU, 8GB) equivalente ao CAX21
  - CX43 (8 vCPU, 16GB) equivalente ao CAX31
- Corrigir location no autoscaler para amd64 (nbg1→fsn1)
- Remover gitlab-pool do autoscaler (substituído pelo Gitea)
- Aumentar worker-pool max para 30
- Corrigir Floating IP e Load Balancer hardcoded em nbg1
- Corrigir stdin em read interativo: usar < /dev/tty e read -p
- Corrigir cleanup de namespace: lidar com Terminating e finalizers
- Corrigir CNPG: nome do deployment, secret e endpointURL duplo
- Aulas afetadas: 08, 09, 10, 11, 12, 14, 17

aula-12/setup.sh

@@ -184,7 +184,27 @@ log_info "=== Instalando Victoria Metrics Stack ==="
 helm repo add vm https://victoriametrics.github.io/helm-charts/ 2>/dev/null || true
 helm repo update vm
 
-kubectl create namespace monitoring 2>/dev/null || true
+# Garantir namespace limpo
+if kubectl get namespace monitoring &>/dev/null; then
+    NS_PHASE=$(kubectl get namespace monitoring -o jsonpath='{.status.phase}' 2>/dev/null)
+    if [[ "$NS_PHASE" == "Terminating" ]]; then
+        log_info "Namespace monitoring em Terminating, limpando finalizers..."
+        # Remover finalizers de CRDs do Victoria Metrics que prendem o namespace
+        for crd in vmagents vmalerts vmsingles vmclusters vmnodes vmrules vmauths; do
+            for resource in $(kubectl get $crd.operator.victoriametrics.com -n monitoring -o name 2>/dev/null); do
+                kubectl patch "$resource" -n monitoring --type=json \
+                    -p='[{"op":"remove","path":"/metadata/finalizers"}]' 2>/dev/null || true
+            done
+        done
+        kubectl wait --for=delete namespace monitoring --timeout=60s 2>/dev/null || {
+            kubectl get namespace monitoring -o json | \
+                python3 -c "import sys,json; d=json.load(sys.stdin); d['metadata'].pop('finalizers',None); json.dump(d,sys.stdout)" | \
+                kubectl replace --raw /api/v1/namespaces/monitoring/finalize -f - 2>/dev/null || true
+            kubectl wait --for=delete namespace monitoring --timeout=30s 2>/dev/null || sleep 5
+        }
+    fi
+fi
+kubectl create namespace monitoring --dry-run=client -o yaml | kubectl apply -f -
 kubectl label namespace monitoring pod-security.kubernetes.io/enforce=privileged --overwrite
 
 if helm status monitoring -n monitoring &> /dev/null; then