add multi model support into the prompt playground so that you can see responses from multiple models in parallel and compare them easily