A Pure Transformer Pretraining Framework on Text-attributed Graphs