feat: merged detection output + parallel model execution

- Multiple models run simultaneously on each frame via ThreadPoolExecutor
- Detections merged with IoU deduplication (same class, high overlap → keep highest conf)
- Single merged_annotated.mp4 output (not per-model videos)
- Single model still uses existing run_pipeline for backward compat
- Active jobs banner on upload page (polls every 5s)
- GPU warning when selecting 3+ models
- 62/62 tests pass

templates/index.html

@@ -257,6 +257,19 @@
   }
   .empty-state p { margin-bottom: var(--space-2); }
 
+  /* Active jobs banner */
+  .active-jobs-banner {
+    display: flex;
+    align-items: center;
+    gap: var(--space-3);
+    padding: var(--space-3) var(--space-4);
+    background: #f0f9ff;
+    border: 1px solid #0EA5E9;
+    border-radius: var(--radius-md);
+    margin-bottom: var(--space-6);
+    font-size: var(--text-sm);
+  }
+
   /* Video delete button */
   .video-delete-btn {
     flex-shrink: 0;
@@ -291,6 +304,12 @@
   <p>Upload a video and select detection models to run analysis.</p>
 </div>
 
+<div class="active-jobs-banner" id="active-jobs-banner" style="display:none">
+  <span class="badge badge-info">&#9654;</span>
+  <span id="active-jobs-text"></span>
+  <a href="/jobs" class="btn btn-ghost btn-sm">View All</a>
+</div>
+
 <form action="/upload" method="post" enctype="multipart/form-data" id="upload-form">
   <input type="hidden" name="video_path" id="video-path-input" value="">