Unsupervised Transformer Pre-Training for Images: Self-Distillation, Mean Teachers, and Random Crops