Add precision and quantization method to VSA models

[mikeheddes]

The current soft_quantize and hard_quantize need to be replaced in favor of a model specific implementation of the precision and quantization for each model.