* [docs] add onnx_infer docs, test=document_fix

* [docs] add python serving docs, test=document_fix

* [docs] add serving cpp docs, test=document_fix

* [TIPC] update serving/python, add keypoint

* [TIPC] update serving/cpp

* [TIPC] add serving_cpp_infer test shell

* [TIPC] remove gpu infer in fleet train, test=document_fix

* [TIPC] add serving python

* [TIPC] alter amp_level=O2