* add kernel for c++ api

* fix compile bugs

* fix kunlun compile bugs

* perfect cmake

* fix compile bugs when run ci-inference

* fix compile bugs

* add non-raw kernel for fluid op

* fix compile bugs

* fix compile bugs

* fix unit test bug

* remove intype arg in cast kernel

* modify conj config in api.yaml by dictionary order

* rm unused code in cast_kernel.cu

* add empty and empty_like kernel in pten

* add empty dev_api

* add pten conj kernel

* modify conj_kernel file path

* add defined cuda macro to cuda/conj_kernel.h

* modify sum mean args

* add GetExpectedPtenKernelArgs for redcue_op

* modify kernel args number

* modify kernel args number

* Reduce reshape kernel functions in pten

* delete notes

* fix bugs when compile

* add variadic_args kernel in pten

* merge develop code

* add variadic_args kernel and benchmark

* change dynamic_cast to static_cast for DeviceContext

* merge the code

* modify code format

* refactor variadic kernel function