Generalized Single-Image-Based Morphing Attack Detection Using Deep Representations from Vision Transformer