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/agents/expenses_agent.py

@@ -1,4 +1,5 @@
 from __future__ import annotations
+import asyncio
 import json
 import logging
 from datetime import date as _date
@@ -133,12 +134,25 @@ class ExpensesAgent(BaseAgent):
                 seen_hashes.add(h)
             unique_receipts.append(r)
 
-        for receipt in unique_receipts:
-            parsed = await self._parse_receipt_text(
-                receipt.get('text', ''), receipt.get('filename', 'receipt'),
+        # Parse all receipts concurrently (bounded by Ollama semaphore)
+        parse_tasks = [
+            self._parse_receipt_text(
+                r.get('text', ''), r.get('filename', 'receipt'),
                 expense_products=expense_products,
-                date_hint=receipt.get('date_from_name'),
+                date_hint=r.get('date_from_name'),
             )
+            for r in unique_receipts
+        ]
+        parsed_list = await asyncio.gather(*parse_tasks, return_exceptions=True)
+
+        for receipt, parsed in zip(unique_receipts, parsed_list):
+            if isinstance(parsed, Exception):
+                logger.warning('expenses_agent: parse failed for %s: %s',
+                               receipt.get('filename'), parsed)
+                parsed = {'vendor': receipt.get('filename', 'Expense'), 'amount': 0.0,
+                          'date': receipt.get('date_from_name') or _date.today().isoformat(),
+                          'product_name': ''}
+
             # Pick product by name match returned from LLM, fall back to default
             product_id = default_product_id
             chosen_name = parsed.get('product_name', '')