MagR: Weight Magnitude Reduction for Enhancing Post-Training Quantization