sync: bring the Space up to the current GitHub tree

fix(gpu): warm LLM into VRAM at startup + health-check grace period — stops the cold-load restart loop

perf(voice): shorter spoken summary (snappier turns) + startup model warmup (no cold-start 5xx)

feat(voice): NotebookLM-style voice UX

Deploy Auralynq RAG (Llama-3.3-70B via HF Inference Providers)