Revitalizing Masked Generative Transformers for Efficient High-Resolution Text-to-Image Synthesis