[IJCV] Show-1: Marrying Pixel and Latent Diffusion Models for Text-to-Video Generation