Low-Rank Correction for Quantized LLMs