Visual Encoders for Data-Efficient Imitation Learning in Modern Video Games