add act grad ops

* add grouup_norm

* fix format. test=develop

* fix xiaodu crash. test=develop

* reduce .so size. test=develop

* compile all targets when LITE_ON_TINY_PUBLISH=OFF

* unordered_map is more convenient when key is customized class

* test=develop

movs unsqueeze frm arm to host.

* test=develop, support WITH_TESTING=ON on windows

test=develop, modify the path of test_file in build_windows.bat

test=develop, modify the docment of windows compilation

test=develop, modify the document

* test=develop, modify the build_windows.bat

fix clang build error

improve sgemm performance on A53

* [NPU] add shape bridge

move shape arm kernel to host

* enhance compare arm kernel

* [NPU] add gather op bridge

* enable reshape arm ut

* [NPU] add lookup_table bridge

[MLU] add some basic support for MLU, including related passes, kernels, gtests and some api in padddle_api.h
Passes：mlu_subgraph_pass ,mlu_postprocess_pass
Kernels:  act，batch_norm, concat, conv, elementwise, fc, interpolate, pool, scale, softmax

* Add hard_swish, ctc_align and reciprocal op, test=develop
* Move some activation ops to extra, test=develop

* rm grad code

* rm grad code

* add mul_grad, test=develop

* add sgd op, test=develop

* test=develop

* add backend

* add kernel

* add grad kernel

* [NPU] subgraph's precision register to kAny

* add mean op, test=develop

* split forward and backward, test=develop

* add lookup_dequant_op

* enhance gather, lookup_table arm kernel uts

* enhance beam_search, beam_search_decoder, increment rigster