fix: 为 search_spu_products 工具注入 user_token 参数

## 问题
即使更新了 Product Agent prompt，LLM 仍然调用 search_products 而非 search_spu_products

## 根本原因
search_spu_products 工具需要 user_token 参数（Mall API 认证必需），
但 product_agent 函数中没有注入此参数，导致工具调用失败或被忽略

## 修改内容

### agent/agents/product.py:169-173
在工具调用前注入 user_token、user_id、account_id 参数：

```python
# Inject context for SPU product search (Mall API)
if result["tool_name"] == "search_spu_products":
    arguments["user_token"] = state.get("user_token")
    arguments["user_id"] = state["user_id"]
    arguments["account_id"] = state["account_id"]
```

## 参数来源
- user_token: 从 Chatwoot webhook 提取（contact.custom_attributes.jwt_token）
- user_id: 从 AgentState 获取
- account_id: 从 AgentState 获取

## 预期效果
LLM 现在可以成功调用 search_spu_products 工具，
返回 Mall API 商品数据并以 Chatwoot cards 格式展示

Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

agent/agents/product.py

@@ -165,12 +165,18 @@ async def product_agent(state: AgentState) -> AgentState:
         
         if action == "call_tool":
             arguments = result.get("arguments", {})
-            
+
+            # Inject context for SPU product search (Mall API)
+            if result["tool_name"] == "search_spu_products":
+                arguments["user_token"] = state.get("user_token")
+                arguments["user_id"] = state["user_id"]
+                arguments["account_id"] = state["account_id"]
+
             # Inject context for recommendation
             if result["tool_name"] == "recommend_products":
                 arguments["user_id"] = state["user_id"]
                 arguments["account_id"] = state["account_id"]
-            
+
             # Inject context for quote
             if result["tool_name"] == "get_quote":
                 arguments["account_id"] = state["account_id"]