Asynchronous prediction handling

Stream: lean-gptf

[Jesse Michael Han]

New feature: gptf will now use tactic.run_async to test the model's predictions simultaneously, which should lead to a faster overall response time.