Add quantization

[seanmor5]

This is important for deploying on mobile/embedded applications. Goal is to achieve feature parity with PyTorch: https://pytorch.org/docs/stable/quantization.html