feat: Enhance inference management with device tracking and telemetry updates

app/overlay/telemetry_overlay.py

@@ -33,6 +33,8 @@ class TelemetryOverlay(IOverlayLayer):
         CPU sys   14.8 %        ← normalised by cpu_count (matches Task Manager)
         CPU core 118.4 %        ← per single core (can exceed 100%)
         Mem        68 MB
+        Inf.dev  mps            ← inference device (only when model loaded)
+        Inf.time  87 ms         ← rolling average of model() call time
     """
 
     def __init__(self) -> None:
@@ -106,4 +108,10 @@ class TelemetryOverlay(IOverlayLayer):
         if snap.memory_mb is not None:
             lines.append(f"Mem      {snap.memory_mb:>5.0f} MB")
 
+        if snap.inference_device is not None:
+            lines.append(f"Inf.dev  {snap.inference_device:>6s}")
+
+        if snap.inference_time_ms is not None:
+            lines.append(f"Inf.time {snap.inference_time_ms:>5.0f} ms")
+
         return lines