[CLI] chat_completion CLI

[yanxi0830]

TL;DR

• adding llama-stack-client inference chat_completion to quickly test a client example w/o getting apps

Test

Non-Streaming

llama-stack-client inference chat_completion --message "hello, what model are you"

Streaming

llama-stack-client inference chat_completion --message "hello, what model are you" --stream

[ashwinb]

chat-completion instead of chat_completion maybe? we should probably standardize with the URLs I think.