[NPU] initial support of asym_int4_rtn

[rnwang04]

Description

1. Why the change?

https://github.com/analytics-zoo/nano/issues/1766#issuecomment-2516410003
Work with https://github.com/intel-analytics/llm.cpp/pull/704 & https://github.com/intel-analytics/llm.cpp/pull/705

2. User API changes

New precision support of asym_int4

3. Summary of the change

• Support new precision asym_int4
• Support running asym_int4 on CPU to verify the correctness
• Support running asym_int4 on NPU with optimize_model=False
• Support running asym_int4 on NPU with optimize_model=True (mp version) for Qwen2-7B
• Support new convert in convert_llm_for_deploy for Qwen2-7B with asym_int4
• Support running asym_int4 on NPU with optimize_model=True (cpp version) for Qwen2-7B
• Support running asym_int4 GW on NPU with optimize_model=True (cpp version) for Qwen2-7B
• Support running asym_int4 split LM head
• Have verified ssave & load of asym_int4 Qwen2-7B

4. How to test?

• Unit test: Please manually trigger the PR Validation here by inputting the PR number (e.g., 1234). And paste your action link here once it has been successfully finished.