cherry-pick some PR about optimize sparse kernel and fix some bug：
#47736 #47703 #47604 #46679 #48439 #49009 #49734

Revert SparseConv support duplicate coordinates

add sync_batch_norm_bn and deliver indices_dict 

新增sparse.is_same_shape、sparse.reshape、sparse.transpose 三个API

cherry-pick : #46322, #46245
Sparse API 支持静态图

cherry-pick : #46016, #46021, #45974

* [Sparse]Sparse add support gpu (#45974)

* [Sparse]Remove unused code (#46021)

* [Sparse] Add infer meta (#46016)

* rename the member function of SparseTensor

* use shorter function names

* fix submanifold conv

* sparse conv support duplicate coordinates

* fix behavior of device_id=None in Tensor.cuda

* fix CI

* sparse support amp

* EagerAmpAutoCasts support sparse

* opt sparse_mask

* add sparse api coalesce

* [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