SwinDANet: leveraging swin transformers with Context-Aware Attention for precise sclera segmentation