Fix: Reorganizar workflow para forzar regeneración de config

- Detener contenedor ANTES de eliminar config.json
- Eliminar certificados y bases de datos viejas
- Generar config.json limpio con OIDC configurado
- Evitar duplicación de steps en el workflow

.gitea/workflows/build-and-deploy.yml

@@ -32,10 +32,25 @@ jobs:
           mkdir -p meshcentral-backup
           mkdir -p meshcentral-config
 
+      - name: Pull latest MeshCentral image
+        run: docker pull ghcr.io/ylianst/meshcentral:latest
+
+      - name: Pull fresh images used in compose
+        run: docker compose pull
+
+      - name: Clean up existing stack
+        run: docker compose --project-name $APP_NAME down
+
+      - name: Remove old config to force regeneration
+        run: |
+          rm -f meshcentral-data/config.json
+          rm -f meshcentral-data/*.db
+          rm -f meshcentral-data/*.crt
+          rm -f meshcentral-data/*.key
+          rm -f meshcentral-data/serverstate.txt
+
       - name: Generate MeshCentral config.json
         run: |
-          # Forzar regeneración del config.json
-          rm -f meshcentral-data/config.json
           cat > meshcentral-data/config.json <<'EOF'
           {
             "settings": {
@@ -92,15 +107,6 @@ jobs:
           chmod -R 755 meshcentral-backup
           chmod -R 755 meshcentral-config
 
-      - name: Pull latest MeshCentral image
-        run: docker pull ghcr.io/ylianst/meshcentral:latest
-
-      - name: Pull fresh images used in compose
-        run: docker compose pull
-
-      - name: Clean up existing stack
-        run: docker compose --project-name $APP_NAME down
-
       - name: Start MeshCentral stack
         run: docker compose --project-name $APP_NAME up -d --remove-orphans --wait