feat: per-class max box fraction (REQ-188)

- labeling/preview/autolabel: max_box_frac + per-class max_box_fracs
  ceiling filter (0=none, 1=off) before NMS, mirrors min_box_frac
- exemplar review-assist: ceiling before max_detections truncation;
  ExemplarLabelRequest + filter panel 'Max box size' slider
- ClassParamsTable: MaxBox column; copy line now REQ-186 amended
  order: class <name> conf <v> iou <v> minbox <v> maxbox <v> container
- docs: design/ui-spec/tasks updated (incl. stale 4-slider narrative)

docs/design.md

@@ -181,7 +181,7 @@ GET    /api/batches/{id}                        # status + review progress (REQ-
 GET    /api/batches/{id}/frames                 # frames + statuses
 POST   /api/batches/{id}/autolabel              # {threshold, class_params} → job (REQ-030,032,034,181)
 POST   /api/batches/{id}/preview                # one frame, run now, nothing written;
-                                                #   +class_params {name: {threshold?, iou_threshold?, min_box_frac?}}
+                                                #   +class_params {name: {threshold?, iou_threshold?, min_box_frac?, max_box_frac?}}
                                                 #     empty override = global value, keys are class names (REQ-181)
                                                 #   +exemplars[] {box:[cx,cy,w,h], positive} — the touched class's own pool only
                                                 #   +exemplar_class_name, +target_class_names; per-class, accumulating
@@ -249,10 +249,10 @@ updated. Range and fps live on the batch, so one video can be used repeatedly.
 **autolabel (REQ-030…034).** Per frame: one `set_image`, then loop each class's prompt (see
 the domain invariants in `../AGENTS.md`), cross-class greedy NMS (REQ-031), write `annotations` rows with
 `source='auto'`. `class_params` (REQ-181) is consulted per class: a numeric
-`threshold` / `iou_threshold` / `min_box_frac` entry replaces the job's global for that class
+`threshold` / `iou_threshold` / `min_box_frac` / `max_box_frac` entry replaces the job's global for that class
 only; classes without an entry — and a run with `class_params` empty or absent — take the
 exact global path. The YOLO confidence floor uses `min([global] + overrides)`, the SAM3
-threshold/dedup/min-box lists are built only from classes that actually override that key,
+threshold/dedup/min-box/max-box lists are built only from classes that actually override that key,
 and `/preview` applies the same overrides so the preview and the job cannot disagree.
 The job and `/preview` also read `project_classes.container` per class when building the
 container-id set for the NMS containment carve-out — same stored flag for both (REQ-184,
@@ -363,7 +363,7 @@ per-class overrides table also carries a **Container** checkbox (REQ-184) that t
 `project_classes.container` through `PATCH /api/projects/{id} { containers: … }`. The same
 table carries a **Copy** button (REQ-186): one click puts every selected class's effective
 settings (per-class override where set, else the global slider; container = the checkbox
-state) on the clipboard as short lines `class <name> conf … iou … minbox … container
+state) on the clipboard as short lines `class <name> conf … iou … minbox … maxbox … container
 true|false`, through `clipboard.js`'s `copyText` (Clipboard API with an `execCommand`
 fallback for insecure contexts) — read-only, no network, no setting changed. Because
 `ClassParamsTable` is shared, the mass modal carries the button too.
@@ -394,7 +394,7 @@ with the shortened pool.
 The run is a **dry run by default** (REQ-175). `ExemplarFilterPanel.jsx` is passed to
 `ReviewSidebar` and rendered above the class list — not over the frame, which is where its
 proposals are drawn. It is mounted only while a run is undecided (`pool.active`): the first
-drag opens it, Apply and Discard close it, and the pool outlives it. Its four sliders
+drag opens it, Apply and Discard close it, and the pool outlives it. Its five sliders
 re-preview on a 250 ms debounce. While a preview is up the canvas **hides the stored shapes
 of the class under review** — the run replaces them wholesale, and leaving them on screen made
 a rejected detection look like it had never gone; other classes stay, dimmed
@@ -419,7 +419,7 @@ POST /api/frames/{id}/exemplar-label
   { "exemplars": [{"box": [x0,y0,x1,y1], "positive": true}, …],   # normalized xyxy
     "class_id": 0,
     "threshold": 0.5, "iou_threshold": 0.8,                       # the panel (REQ-175)
-    "min_box_frac": 0.002, "max_detections": 100,
+    "min_box_frac": 0.002, "max_box_frac": 1.0, "max_detections": 100,
     "apply": false }
   → { "shapes": [{"geometry": …, "score": …, "source": "manual"|"auto"}, …],
       "applied": false, "redetected": true, "message": null,
@@ -440,7 +440,7 @@ then rewrites the frame **for that class only**:
   while the drawn shapes survive.
 
 The panel's filters run before any of that, in the order the batch job uses them: area floor,
-then NMS (`labeling.deduplicate`), then the cap on how many survive.
+then ceiling (REQ-188), then NMS (`labeling.deduplicate`), then the cap on how many survive.
 
 The delete-and-reinsert happens in one `db.cursor()` transaction, so the frame is never
 briefly empty. Other classes on the frame are never touched. The GPU lock is taken with a