* move elementwise_sub and elementwise_sub_grad XPU kernel to PHI,test=kunlun

* modify code style,test=kunlun

* modify elementwise_subtract_grad_kernel.cc,test=kunlun

* modify elementwise_subtract_kernel.cc,test=kunlun

* modify elementwise_subtract_grad_kernel.cc,test=kunlun

* modify elementwise_kernel.cc and elementwise_subtract_kernel.cc,test=kunlun

* modify codestyle,test=kunlun

* modify elementwise_kernel.cc,test=kunlun