infer deploy:fix trt keypointbatchsize; lite deploy: update total pipeline time from per person to per image (#4339)

* add fairmot deploy

* separate main_jde

* update copyright & add PrintBenchmark

* refine lap

* update picodet inference deploy

* add keypoint top-down cpp infer; total:5ms,pre:0.7ms,infer:3.8ms,post:0.01ms

* add dark_pose support

Revert "add keypoint top-down cpp infer; total:5ms,pre:0.7ms,infer:3.8ms,post:0.01ms (#4058)" (#4216)

This reverts commit 08619b4e.

* fix not find dirent.h on windows

* using cv::glob replace fileststem, fileststem need std++17

* fix bug when benchmark=True

* split warmup and repeat

* support xpu inference

* Automatically set use_dynamic_shape

* fix image_shape in export_model

* fix batch_size when trt infer

* bs>1 for YOLO

* update doc, test=document_fix
Co-authored-by: NGuanghua Yu <742925032@qq.com>

* s2anet fix export and deploy

* remove redundant code

* fix cpp for s2anet deploy

* fix bug of get_categories

* rename poly_to_rbox to poly2rbox

* add some comment for function

* add inference benchmark

* add ext_op build doc, test=fix_document

* add note for JetPack version with machine

* fix path of script, test=document_fix

* update for ci-travis, test=document_fix

* add benchmark, test=document_fix

* update style of number, test=document_fix

* add difference of dynamic shape and fix shape

* fix paddle inference name problem, and update doc

* set default value

* fix inference lib name in static

* update deploy doc, fix build error of windows

* add annotation of PADDLE_LIB_NAME

* set min_subgraph_size=33 as default, fix bug of fcos_dcn_r50_fpn_1x

* Update install doc and deploy doc.

* fix doc link, test=dygraph, test=document