* PTQ save quantized model

* Wrap simulated layer

* post process the inference model

* Add wrap for functional api
* Refine the wraped api
* Add unit test for quant functional layers
* Update all unit tests for dygraph qat