fix: increase timeout and parallelize receipt processing

- ab_ai_bot: raise requests.post timeout 120s -> 600s so long OCR+LLM
  runs don't silently drop the reply in Discuss
- upload: run parse_upload in ThreadPoolExecutor so tesseract OCR
  doesn't block the FastAPI event loop
- expenses_agent: parse all receipts concurrently with asyncio.gather
  (Ollama semaphore caps parallelism at 2); reduces 13-receipt LLM
  time from ~39s sequential to ~20s parallel

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

agent_service/routers/upload.py

@@ -31,12 +31,18 @@ async def upload(
         raise HTTPException(status_code=status.HTTP_503_SERVICE_UNAVAILABLE,
                             detail='Agent service not ready')
 
+    import asyncio
+    from concurrent.futures import ThreadPoolExecutor
+    _ocr_executor = ThreadPoolExecutor(max_workers=2)
+
     receipts: list[dict] = []
+    loop = asyncio.get_event_loop()
     for f in files:
         data = await f.read()
         filename = f.filename or 'receipt'
         try:
-            parsed = parse_upload(filename, data)
+            # parse_upload may run OCR (CPU-bound) — offload to thread pool
+            parsed = await loop.run_in_executor(_ocr_executor, parse_upload, filename, data)
             receipts.extend(parsed)
             logger.info('upload: parsed %s → %d receipt(s)', filename, len(parsed))
         except Exception as exc: