[alerts] Fix status-page-pusher crash + Prometheus backup push

## status-page-pusher (ExternalAccessDivergence false positive)
The pusher was crashing with `AttributeError: 'list' object has no attribute
'get'` at line 122 — the uptime-kuma-api library changed the heartbeats return
format. Fixed by making beat flattening more robust: handle any nesting of
lists/dicts in the heartbeat data, and add isinstance check before calling
`.get()` on the latest beat.

## Prometheus backup (PrometheusBackupNeverRun)
The backup sidecar's Pushgateway push was silently failing because `wget
--post-file=-` needs `--header="Content-Type: text/plain"` for Pushgateway
to accept the Prometheus exposition format. Added the header. Also manually
pushed the metric to clear the `absent()` alert immediately.

Note: ExternalAccessDivergence still fires because 5 services (ollama, pdf,
poison, dns, travel) ARE genuinely externally unreachable but internally up.
This is a real issue (likely Cloudflare tunnel routing) not a false positive.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

stacks/monitoring/modules/monitoring/prometheus_chart_values.tpl

@@ -246,7 +246,7 @@ server:
             ls -t /backup/prometheus_*.tar.gz 2>/dev/null | tail -n +3 | xargs rm -f 2>/dev/null
 
             # Push success metric to Pushgateway for alerting
-            echo "prometheus_backup_last_success_timestamp $(date +%s)" | wget -qO- --post-file=- http://prometheus-prometheus-pushgateway.monitoring:9091/metrics/job/prometheus-backup 2>/dev/null
+            printf "prometheus_backup_last_success_timestamp %s\n" "$(date +%s)" | wget -qO- --header="Content-Type: text/plain" --post-file=- http://prometheus-prometheus-pushgateway.monitoring:9091/metrics/job/prometheus-backup 2>/dev/null
 
             echo "$(date) Backup complete. Files in /backup:"
             ls -lh /backup/prometheus_*.tar.gz 2>/dev/null || echo "  (none)"