* Revert "Revert "make inference_api_test compile with dynamic linking library (#41944)" (#43882)"

This reverts commit e6d81ddf.

* modify third_party cmake

* move SKIP_CPP_TEST return in the begining

This reverts commit f8b9073f.

* make inference_api_test use import_lib + dll to smaller its size;test=document_fix;test=windows_ci_inference

* make place.h not export symbols; change the order of inference test dependencies

* cant delete analysis pass ..in inference_analysis_test or build test_analyzer.exe will failed; temporarily comment test need paddle_inference_c

* add onnx path; cp onnxruntime.dll to tests/api

* make c_ut depend on paddle_inference_c_shared rather than paddle_inference_c

* make c_ut depend .dll in windows and .lib in linux;path add capi_exp

* init context pool by analysis predictor

* make windows and linux both use shared; compile shared even on_infer off

* move case:profile after case:fuse_statis to avoid context pool null problem

* make c_ut depend .dll in windows and .lib in linux;path add capi_exp

* open inference_ut;test=windows_ci_inference

* inference_ut need onnx;test=windows_ci_inference

* disable trt_split_converter_test; use higher parallel level

* too high parallel will cause ut timeout

* run all demo ci before exit;test=document_fix;test=windows_ci_inference

* fix bug;test=document_fix;test=windows_ci_inference

* improve log

* commetn test code

* modify according to zhouwei's comments

* set scipy and numpy version suit for py3.6

* pynacl1.5.0 which is needed by PyGithub built failed in python36, change it to 1.4.0 also works;test=document_fix;test=windows_ci

* np.corrcoef support parameter since 1.20

* delete test code

* set auto to reduce core_avx/noavx.pyd size

* set CUDA_ARCH_NAME in each case

* fix windows compile test case error

* test windows ci

* cmake add onnxruntime

* cmake add onnxruntime

* test windows ci

* auto_code_generator add ort lib copy

* fallback modify windows ci bat

* ci notest;test=document_fix;test=windows_ci_inference;test=windows_ci;test=windows_op

* change inference demo_test build method to ninja to choose visual studio version automaticly

* notest;test=windows_ci_inference

* set cuda of demo_ci by arg,fix bug of ninja compile,test=document_fix;test=windows_ci;test=windows_ci_inference

* fix bug;test=document_fix;test=windows_ci;test=windows_ci_inference

* fix bug;test=document_fix;test=windows_ci_inference"

* set lib_path according to generator

new cuda arch compile method;test=document_fix;test=windows_op;test=windows_ci_inference;test=windows_ci (#41259)

* add CUDA_TOOLKIT_ROOT_DIR option in cmake command

* make sure dir in cmake use / rather than \

* make vcvars64 and cuda_version can be set in xly pipe

* make third_party_path reused by ci and build pipe;test=windows_ci_inference;test=windows_op;test=windows_ci

* change CUDA implementaion of uniform/gaussian OP

* fix unittest

* force close eager_generator.exe

* modify according to zhouwei's comment

* support vs2019 compilation in windows

* not modify pow_op's original compute logic

* add trt_mobilenet demo test

* only deliver trt_root in run.sh

* modify for wincheck-inference case

* modify according to zhouwei's comment

* open with_trt and block failed unittests in windows

* test

* test on_infer off problem in windows

* turn off on infer in windows-ci

* for pure fp16

* opt topk

* modify for wincheck-inference case

* modify according to zhouwei's comment

* modify according to zhouwei's comment 2nd time
Co-authored-by: Nzhangkaihuo <zhangkaihuo@baidu.com>