Show HN: HyperSAE – Hyperbolic Sparse Autoencoders for LLM Interpretability