Enabling Instructional Image Editing with In-Context Generation in Large Scale Diffusion Transformer