refactor(upscale): drop LoRA support per spec

fix(upscale): crop upscaled image to multiple-of-16 dims before refinement

fix(upscale): pass post-upscale dims to pipe + LoRA label hints model compat

feat(progress): gradio progress(track_tqdm=True) on all three handlers

fix: pool-stashed transformer swap + MPS-safe vram + corrected model-zoo anchor

fix: post-review polish — slider defaults, error handling, comment

ci: ruff + pytest on push/pr (l1+l2, no gpu deps)

feat(modes): upscale handler (realesrgan + z-image-turbo refinement)

feat(modes): controlnet handler (turbo + union 2.1 + preprocessor)

feat(modes): t2i handler (base + turbo) with transformer swap and lora ctx