fix: add verbose deploy diagnostics

Enable strict shell tracing and explicit progress logs in the EC2 deploy block so container replacement failures are visible and easier to debug.

.github/workflows/deploy-ec2.yml

@@ -71,7 +71,7 @@ jobs:
       - name: Deploy on EC2 with docker compose
         run: |
           ssh -i ssh_key.pem -o StrictHostKeyChecking=no ${{ secrets.EC2_USER }}@${{ secrets.EC2_HOST }} << 'ENDSSH'
-          set -e
+          set -euxo pipefail
           cd ${{ env.REMOTE_DIR }}
           echo "=== BEFORE DEPLOY ==="
           docker compose -f docker-compose.ec2.yml --env-file .env ps || true
@@ -79,10 +79,13 @@ jobs:
 
           docker load -i junohub-app_${{ env.IMAGE_TAG }}.tar
           rm -f junohub-app_${{ env.IMAGE_TAG }}.tar
+          echo "Loaded new image tar successfully"
 
           echo "=== START DB / ENSURE EXTENSION ==="
           docker compose -f docker-compose.ec2.yml --env-file .env up -d --remove-orphans db
+          echo "DB up command completed"
           docker compose -f docker-compose.ec2.yml --env-file .env exec -T db sh -lc "psql -U \"\${POSTGRES_USER:-postgres}\" -d \"\${POSTGRES_DB:-dropshipping}\" -c \"CREATE EXTENSION IF NOT EXISTS vector;\""
+          echo "Vector extension check completed"
 
           # Stop and replace app container only (no volume removal)
           echo "=== REPLACE APP CONTAINER ==="
@@ -91,6 +94,7 @@ jobs:
           # Fallback in case old container is detached from compose state but still running with the canonical name.
           docker rm -f juno-app-1 2>/dev/null || true
           docker compose -f docker-compose.ec2.yml --env-file .env up -d --no-deps --remove-orphans app
+          echo "App replace command completed"
 
           echo "=== AFTER DEPLOY ==="
           docker compose -f docker-compose.ec2.yml --env-file .env ps