* open header for custom kernel

* add core utils

* tidy core code

* tify header

* tidy include

* tidy namespace

* resolve conflit

* fix unittest and coverage

* remove platform using

* resolve conflict

* resolve conflict

* fix digamma namespace error

* fix xpu full kernel error

* fix xpu full kernel error

* polish details

* add place for lib storage

* cast xpu kernel init

* cast xpu kernel

* replace with raw cast xpu kernel

* fix cast kernel bug

* add the missing break

* modify namespace and header file