EzAudio: Enhancing Text-to-Audio Generation with Efficient Diffusion Transformer