Client thread couldn't render properly when using OpenAI chat completion API

[sq-yuan]

Currently the SDK uses OpenAI response API's response.item_id as the thread item id, instead of generating it explicitly.

However, with chat completion API, the Agent SDK currently populates a fake_id. Because of which, all assistant messages are using fake_id. And from client side, it'll just override one another.

I can create some example for this issue if needed.