fix(renderer): use exact plot_videos.py scaling pipeline (* 3 * 240 + offset)

fix(hq): correct config loading, auto-scaling rendering, and dynamic watermarks

feat: integrated cloud-based diffusion inference module