Imitation Learning from Vague Feedback