[runtime] Benchmark streaming Triton TRT-LLM first chunk latency (#188)

Feat: add triton runtime decoupled spark_tts python backend + decoupled tensorrt_llm backend (#126)

update readme for Triton

Merge pull request #92 from yuekaizhang/triton

change usage

add license

Merge branch 'SparkAudio:main' into triton

add code commit

update docker file

update http client; launch script

Merge pull request #54 from xcv58/macos

clean codes

space clean up

Update webui.py as well

Add support for macOS Metal Performance Shaders (MPS) and CPU fallback