Location-Aware Pretraining for Medical Difference Visual Question Answering