Fix blank TPS: read tick timing from vanilla 'tick query' instead of spark

spark sends command output asynchronously, so its RCON responses are
empty and the sampler recorded null TPS/MSPT. Use the synchronous
vanilla 'tick query' as the primary source (treating paused/frozen idle
servers as healthy), keep spark as fallback, and route spark ping /
healthreport through the server console (mc-send-to-console) with the
log watcher capturing ping lines. Lag endpoint now falls back to the
freshest recorded ping sample.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

dashboard/src/app/api/lag/route.ts

@@ -40,7 +40,13 @@ export async function GET() {
 
     const serverHealthy = (tps === null || tps >= 18) && (mspt === null || mspt <= 40);
     const players = online.map((name) => {
-      const ping = livePings[name] ?? null;
+      // spark's RCON reply is often empty (async output) — fall back to the
+      // most recent logged sample if it's fresh enough.
+      const hist = historyByPlayer[name] ?? [];
+      const lastSample = hist.length > 0 ? hist[hist.length - 1] : null;
+      const ping =
+        livePings[name] ??
+        (lastSample && Date.now() - lastSample.ts < 5 * 60 * 1000 ? lastSample.ping : null);
       return {
         name,
         ping,