Text this: Optimizing deep learning models for glaucoma screening with vision transformers for resource efficiency and the pie augmentation method.