Integrating quantization aware training to SparseXP.

[madlag]

Thanks, the addition of quantization aware training to the set of available model optimizations is definitely an exciting feature!