Support group router for moe models

[RunningLeon]

Motivation

Support group router for moe models

Use cases (Optional)

pipeline

from lmdeploy import pipeline, GenerationConfig, PytorchEngineConfig

if __name__ == '__main__':
    backend_config = PytorchEngineConfig(hf_overrides=dict(router_n_groups=4))
    model_path = 'Qwen/Qwen3-30B-A3B'
    pipe = pipeline(model_path, backend_config=backend_config)

    resps = pipe(['Hi.'])
    for res in resps:
        print(res)

api server

lmdeploy serve api_server \
Qwen/Qwen3-30B-A3B \
--backend pytorch \
--hf-overrides '{"router_n_groups": 4}'

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit tests to ensure the correctness.
3. If the modification has a dependency on downstream projects of a newer version, this PR should be tested with all supported versions of downstream projects.
4. The documentation has been modified accordingly, like docstring or example tutorials.