The Cambrian Explosion of Mixed-Precision Matrix Multiplication for Quantized Deep Learning Inference