* [XPU] add fp16 support for compare ops.

* fix ci.

* Replace paddle::experimental::DataType as phi::DataType

* restore custom_device.cc

* Check structed kernel for new executor static build

* Update code

* Ready for resnet50

* Move transfer_dtype to phi

* Ready for transformer

* Fix CI errors

* Fix layer_norm InferMeta

* Remove layer_norm infermeta fix

* standard api

* fix sparse bugs

* fix xpu bugs, test=kunlun

* remove hard code for custom unittest

* open ci, test=kunlun

* deal with conflict

* migrate compare op(E/NE/LE/LT/GE/GT) from fluid to phi

* fix compile error

* fix memory alloc; test=kunlun

* refine code; test=kunlun

* replace PADDLE_ENFORCE with PADDLE_ENFORCE_XDNN_SUCCESS