Image as a Foreign Language: BEiT Pretraining for Vision and Vision-Language Tasks