Style Description based Text-to-Speech with Conditional Prosodic Layer Normalization based Diffusion GAN