QuantEase: Optimization-based Quantization for Language Models