Support GLM4, ChatGLM and Baichuan models

[JoeZhang-0x000]

Background

InfiniLM needs support for additional model families: GLM4, ChatGLM, and Baichuan.

Requested support

• Register GLM4, ChatGLM, and Baichuan model types in config/model factory paths.
• Add causal LM wrappers for these model families.
• Handle Baichuan/ChatGLM tokenizer and chat prompt behavior in the example inference flow.

Expected outcome

Users can load and run these model families through the existing InfiniLM inference path.