Add MCP control server, live API-generation view, and repetition penalty

- scripts/jwash_mcp.py: MCP server (FastMCP/stdio), an HTTP client of the running
  J-Wash server so an external LLM can drive an already-loaded model. Tools:
  generate, find_token, list_layers, scale_token/replace_token (pure-weights,
  layers required), set_intensity, list_edits/reset_edits.
- api/app.py + ui: /api/generate now records the last exchange (surfaced in
  /api/status) and broadcasts on /ws when done; the UI shows a "generated via
  API/MCP" panel at the top of the chat and an Options "API monitor" toggle that
  swaps the 2s status poll for an event-driven refresh.
- sampling: repetition penalty (default 1.0, applied in model_manager._sample),
  exposed as a "rep" field in the chat controls; intentionally not exposed
  through the MCP.
- ui: remove the redundant "md" chat toggle (already available in Options).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

api/app.py

@@ -88,6 +88,9 @@ fit_manager.on_progress = _broadcast_fit
 # concurrent HF downloads: one state per repo_id
 _downloads = {}
 _downloads_lock = threading.Lock()
+# last synchronous /api/generate exchange (CLI/MCP): surfaced in /api/status so
+# the UI can show what an external client generates, without persisting it
+_last_generation = None
 
 
 class LoadRequest(BaseModel):
@@ -203,6 +206,7 @@ def api_status():
         "interventions": interventions.summary(),
         "interventions_scale": interventions.global_scale,
         "interventions_mode": interventions.mode,
+        "last_generation": _last_generation,
     }
 
 
@@ -740,6 +744,23 @@ async def api_generate_sync(req: GenerateSyncRequest):
         raise HTTPException(500, str(exc))
     if done.get("error"):
         raise HTTPException(500, done["error"])
+    global _last_generation
+    _last_generation = {
+        "n": (_last_generation or {}).get("n", 0) + 1,
+        "prompt": next(
+            (m.get("content", "") for m in reversed(req.messages) if m.get("role") == "user"),
+            "",
+        ),
+        "text": done.get("text", ""),
+        "stats": done.get("stats"),
+    }
+    # nudge any watching UI to refresh once this API generation is done (used by
+    # the "API monitor" mode, which drops the 2s status poll for event-driven refresh)
+    for ws in list(_ws_locks):
+        try:
+            await _ws_send(ws, json.dumps({"type": "api_generation"}))
+        except Exception:
+            pass
     return {"text": done.get("text", ""), "stats": done.get("stats")}