perf(ui): probe sources asynchronously so /api/status no longer blocks on upstream reachability

ensureProbe() called Registry.ProbeAll() synchronously on every status request when >30s had passed since the last probe, blocking the response on the slowest upstream (zen ~1s, frank ~1.9s, total ~1.5s) — this stalled WebUI page loads and tab switches. Now the probe runs in a background goroutine (15s budget); status returns in ~7ms and the 3s stats poll picks up updated reachability next tick. Deployed (bak .bak.20260811s), server active.

internal/gateway/server.go

@@ -377,6 +377,9 @@ func (g *Gateway) handleModels(w http.ResponseWriter, r *http.Request) {
 }
 
 // ensureProbe triggers a live source probe at most once every 30s.
+// The probe runs asynchronously so a slow/stuck upstream never blocks the
+// status response — reachability data is eventually-consistent and the UI
+// polls /api/stats every 3s anyway, so the next refresh picks it up.
 func (g *Gateway) ensureProbe(ctx context.Context) {
 	g.probeMu.Lock()
 	due := time.Since(g.lastProbe) > 30*time.Second
@@ -384,9 +387,14 @@ func (g *Gateway) ensureProbe(ctx context.Context) {
 		g.lastProbe = time.Now()
 	}
 	g.probeMu.Unlock()
-	if due {
+	if !due {
-		g.core.Registry().ProbeAll(ctx)
+		return
 	}
+	go func() {
+		probeCtx, cancel := context.WithTimeout(context.Background(), 15*time.Second)
+		defer cancel()
+		g.core.Registry().ProbeAll(probeCtx)
+	}()
 }
 
 // handleResetHealth (admin) clears the per-source backoff state so a fixed