[cherry pick] Support optional residual add in fused ops and slice large tensor for cudnn_softmax (#43719)

 [cherry pick] Support optional residual add in fused ops and slice large tensor for cudnn_softmax

cherry-pick #43635 #43681 #43474

* Optimize the computation of log_softmax

* modify the var name

* Optimize performance of log_softmax

* delete unity build

* modify to phi

* fix

* fixfixfixfix

* fix

* fix

* fix

* fix

* simplify

* fix

* fix enforce

* add softmax log_softmax

* refine rocm

* refine unittest

* support cudnn kernel moving

* polish cmake rules

* add unittest for coverage

* remove orig kernel

* remove softmax cudnn kernel

* fix softmax test failed

* fix npu func error

* resolve conflict

* rename gpu dnn kernels

* fix name rule error

* fix compile error

* update fp16 namespace

* Optimize performance of softmax_bwd when axis!=-1

* fix

* fix

* fix

* fix

* Optimize performence of softmax_fwd when axis!=-1

* use functor

* support hip

* fix functor

* refine structure for cuda and rocm

* update

* update

* update

* update

* softmax backward optimize

* optimize softmax forward

* vec softmax fw

* vec softmax bw

* add a message argument for compiler compatibility