* sparse support amp

* EagerAmpAutoCasts support sparse

* opt sparse_mask

* add sparse api coalesce

* clean glog header in public header

* move marco pos

* [Sparse] add SparseCsrTensor fused_attention kernel and API

* fix comment

* [Sparse]add SparseTensor mv kernel(csr*dense_vec->dence_vec, coo*dense_vec->dense_vec)

* fix CI

* perfect copy

* deal with conflict

* deal with conflict

* fix compile bugs

* fix unittest bugs

* change code format

* deal with conflict

* modify code by review

* fix ce bugs

* fix ce bugs

* add lo

* perfect code format

* deal with conflicts

resort .cu headers, set clang-format not sort include block and consider .cu as main source file (#43633)

* add new API/OP(csr->csr) of SparseTensor softmax

* fix comment

* fix to_sparse_csr

* add some kernel(csr*dense->csr, dense*dense->csr) of SparseTensor matmul

* fix CI

* fix CI

* fix comment

* fix comment

* use GpuMemcpy and GpuMemset

* sparse convert kernel support double dispatch by indices dtype

* cudaMemcpyKind->gpuMemcpyKind

* remove storage

* add glog include

* add glog include

* add glog include

* remove shared_storage

* fix bug

* fix rnn bug

* chnage dispatch to visit

* resolve conflict