[speechx] high performance inference for ds2

[speechx] refactor speechx egs

[tts]update synthesize_e2e.sh

[speechx]add reset of decoder&frontend

[Audio] fix setup.py of paddleaudio

[Doc] Update demo doc

[Audio][Feature][Unittest][Benchmark]Add unittest and benchmark for feature apis.

[test] fix benchmark

[speechx] Add run.sh for example

[speechx]align nnet & decoder

Revert "align nnet decoder & refactor"

align nnet decoder & refactor

[test][cli] update cli test case

[server] add server cls

[speechx]add raw_audio & feature_cache