Add server-side health checks for service online status

- update-health.js: pings every service URL from script.js (single
  source of truth) in parallel every minute via cron and writes
  health.json; any HTTP response below 500 counts as online, 5xx as
  error, timeouts/DNS failures as offline; TLS validity ignored
- Card badges now show real state instead of a static Online label:
  grey Checking, green Online (with latency tooltip), amber Error
  (HTTP code), red Offline (reason); refreshed every 30 s
- Entrypoint runs a first sweep at boot; nginx serves /health.json
  with no-cache; README documents the behavior

nginx.conf

@@ -13,6 +13,11 @@ server {
         add_header Cache-Control "no-cache" always;
     }
 
+    # Service health checks; refreshed every minute by cron.
+    location = /health.json {
+        add_header Cache-Control "no-cache" always;
+    }
+
     # Open WebUI chat proxy.
     location /api/chat {
         proxy_pass         http://127.0.0.1:3001;