ci: switch deploy+smoke jobs to self-hosted runner (peptide-vps)

Route deploy-staging, deploy-production, and smoke jobs through the
org self-hosted runner [self-hosted, peptide-vps] on KVM8 VPS.

This fixes the Hostinger SSH throttle: Azure eastus2/ncus ranges are
blocked at :65002; the VPS has a stable, allowlisted IP.

Callers' CI/validate jobs remain on ubuntu-latest (GitHub-hosted).

Bumps to v1.7.1.

Agent-Session: deploy-self-hosted-20260617

.github/workflows/deploy-app.yml

@@ -35,7 +35,7 @@ permissions:
 jobs:
   deploy-staging:
     name: Deploy to Staging
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, peptide-vps]
     environment: staging
     steps:
       - uses: actions/checkout@v4
@@ -83,7 +83,7 @@ jobs:
     # skipped smoke (run_smoke: false) still allows prod via the if-condition.
     needs: [deploy-staging, smoke]
     if: ${{ !cancelled() && needs.deploy-staging.result == 'success' && needs.smoke.result != 'failure' }}
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, peptide-vps]
     environment: production
     steps:
       - uses: actions/checkout@v4

.github/workflows/smoke.yml

@@ -32,7 +32,7 @@ on:
 jobs:
   smoke:
     name: Playwright smoke suite
-    runs-on: ubuntu-latest
+    runs-on: [self-hosted, peptide-vps]
     timeout-minutes: 15
 
     steps: