Fix llama.cpp CPU deployment on Spaces

Fix model loading: verbose logging, proper GPU detection, detailed error messages

Remove show_api (removed in Gradio 6) from launch kwargs

Remove deprecated show_copy_button param from Chatbot (Gradio 6)

Migrate to Gradio 6 API: theme/css/head to launch(), remove Chatbot type param

Add ZeroGPU support: @spaces.GPU + n_gpu_layers=-1

Case Lantern: GGUF backend + custom dark frontend for Build Small Hackathon