* Remove the output for moving_average_abs_max_scale op, test=develop

* fix div zero test=develop

* fix div zero test=develop

* add hostdevice function test=develop

* add eps when is zero test=develop

TemporaryAllocator is a singleton used for allocating memory for Cudnn. Since it is a singleton, we can delete it for better performance in memory.

We replace TemporaryAllocator by CUDADeviceContextAllocator and CUDADeviceContextAllocation, which uses stream callback to delete the memory allocated for the stream to avoid singleton.

Also added data_feed_proto to operator to fix CI in CPU compilation

* add quant_dequant_moving_avg_max_abs op
test=develop

* add more note for quantdequant op
test=develop

* Add MovingAverageAbsMaxScale operator which is only used for calculating the quantization scale.

* test=develop

* change the output into inplace. test=develop

* Revert "test=develop"

This reverts commit 696cf62699ba1e1c98f61f7345ac7060010eb29a.

* Revert "change the output into inplace. test=develop"

This reverts commit a19acd20f07eee82622701a3015e6e9c073a5e0b.

* test=develop.

* update the MovingAverageAbsMaxScaleOp test. test=develop

* Add moving average absmax op in quantilize-aware training.

* Improve and fix fake_quantize_op.

* Add a fake_quantize_op, which quantize an input tensor to a tensor with lower bits.