test: 清理 backlog——chat_structured 超限检测 + client 上下文管理器

src/genesis/inference/client.py

@@ -48,6 +48,13 @@ class HttpLLMClient:
         self._retry_backoff = retry_backoff
         self._client = httpx.Client(timeout=timeout_sec, transport=transport)
 
+    def __enter__(self) -> HttpLLMClient:
+        """支持 with 块：退出时自动关闭底层连接。"""
+        return self
+
+    def __exit__(self, exc_type, exc_val, exc_tb) -> None:
+        self._client.close()
+
     def chat(
         self,
         *,