Theoretically Better and Numerically Faster Distributed Optimization with Smoothness-Aware Quantization Techniques