* Add unitest from shilong

* Add kernel code from shilong

* fix codestyle

* add broadcast_shape test

* fix unitest

* fix unitests

* fix unitest

* add 0D grad support

* add 0D grad support

* add 0D grad support

* fix 0D tensor

* fix 0D

* fix xpu 0D

* fix expand kernel

* fix xpu expand

* Fix 0D kernel

* fix 0D

* fix 0D

* fix 0D

* fix 0D

* fix XPU top_k

* cancel the modify of xpu

* add XPU 0D tensor

* fix 0D