feat(dipu): support ForceFallbackToCPU (#929)

* feat: support ForceFallbackToCPU

* fix: use bool instead of string

* feat: init out as empty if sizes not equal

* fix: throw error when out is cpu tensor

* fix: replace list and tuple for Python3.8

* doc: add comments

* refactor: rename enable_fallback_to_cpu

dipu/SupportedDiopiFunctions.txt

@@ -44,7 +44,6 @@ diopiBitwiseOrInp
 diopiBitwiseOrInpScalar
 diopiBitwiseOrScalar
 diopiBmm
-diopiCastDtype
 diopiCat
 diopiCdist
 diopiCdistBackward