provide intel optimized models for pytorch and mxnet

pytorch:
a. Support resnet50 model inference
b. Support calibration
c. Support both int8 and fp32
mxnet:
a. Support wide&deep model training and inference
b. Support accuracy and performance evaluation in inference benchmark
c. Support calibration
d. Support int8 and fp32