add search_group_padding op and x86 kernel

* [X86][CUDA] add sequence_arithmetic op , x86 kernel, cuda kernel and unit test

test=develop

* add sequence_arithmetic cuda kernel unit test

test=develop

* add cuda match_matrix_tensor op and test, test=develop

* add search_seq_depadding x86 and cuda
* add match_matrix_tensor x86
* add search_grnn x86, no test

fix the cuda bilinear and nearest precision caused by data type conversion. 

- add var_conv_2d op

- add var_conv_2d x86 kernel

- add var_conv_2d x86 test

* fix the compile error for cuda and fix unit test

* fix conv kernel bugs and open mobilenet ci test=develop

* update lookup_table arm x86, test=develop

* add lookup_table_v2_op for compatibility, test=develop

- add sequence_reverse op

- add sequence_reverse kernel for x86 and cuda

- add sequence_reverse_test for x86 and cuda

* add cast op for x86 platform

* alter the struct to class to hide the data and alter the pointer

- add sequence_concat op

- add sequence_concat kernel for x86 and cuda

- add sequence_concat_test for x86 and cuda

align the lite nearest， bilinear op to fluid on arm and cuda

* update concat and unsqueeze, test=develop

add cuda kernel:lookup table

* fix pooling bug and speed

* fix build error

* delete VLOG in pool, test=develop

* add openmp, test=develop

* fix lite/kernels/arm/pool_compute_test basic_pooling compute error bug, test=develop

* move multiclass_nms kernel back to host test=develop

* move layer_norm OP and arm_kernel into extra type since it's added after release/v2.0-beta1 and not related with CV test=develop

* fix code_style test=develop

[LITE][NPU] Add fusion_elementwise_add_activation bridge, remove the limitation of the dimensions of input tensors in the graph compute kernel, and refine log message (#2395)

test=develop

mv deviceinfo.workspace and other relative member to thread_local_storage

We have changed 11 arm_kernels into extra type in #2347 , which has caused test_compiling failure. In this PR , we move their 11 related arm_kernel_test into build_extra=ON

update reshape op to support multiple input types of shape.
priority: input(ShapeTensor) > input(Shape) > attr(shape)

update slice op to support multiple iput types of starts and ends.
priority: input(StartsTensor) > input(StartsTensorList) > attr(starts)

fill_constant kernel only registered float type, only the float data type is produced, which is obviously a bug.

Now, produce data based on the data type attr.

By the way, fix the cast kernel bug.

* fix step rnn model run bugs test=develop

* move some basic ops into extra type to reduce library size test=develop (#2347)

* [XPU] add batchnorm op bridge and unit test

test=develop

* fix DMLC_USE_GLOG

test=develop

test=develop

* add tanh op in x86 platform and its unittest, test=develop

* add gelu op on x86 platform and add its unittests, test=develop

* update depends for math_function for activation for gelu, test=develop

test=develop

* Add supporting for Huawei offical DDK
* Fix the param of graph op in NPU graph computing kernel

test=develop

* [XPU] add mul op bridge and unit test

test=develop

* use tmp tensor for transposed y

test=develop

test=develop

* Initial support for XPU
* Fix compiling errors of XPU
* Move XPU op kernel bridges from backends to kernels to fix deps order
* Change the namespace and directory of XPU bridges
* Add XPU SDK
* Fix header files and namespace of XPU SDK
* Add unit tests for relu and conv2d ops
* Restore the modification of paddle_api_test
* Supports simple model which contains only a relu layer
* Add compiling scripts for XPU
* Fix compiling errors of XPU
* Add comments for XPU LoadModel and BuildModel

* make inceptionv4, resnet50, googlenet can run on x86 paltform and fix the compare part in x86 unittests, test=develop

* fix googlenet tests for benchmark record, test=develop

* [framework][profile] fix profile dump bug when op is feed and fetch test=develop (sangoly)

* modify yolobox_cuda to support multiple runs test=develop