fix(godcmd): answer when #model is given more than one argument - #3305

What

#model (#模型) handled no arguments and exactly one argument, and nothing else, so #model gpt-4o now matched no branch and left the internal result = "string" placeholder (plugins/godcmd/godcmd.py:262) as the reply body -- the bot sent the bare word string back to the user. The branch now answers a wrong argument count with usage text, 请提供一个模型名称, the way its two one-argument siblings already do (set_openai_api_key line 305, set_gpt_model line 319). Zero-argument and single-argument behaviour is untouched, including the existing 模型名称不存在 answer.

Fixes #3304

Checks

• RED on the current tip (86718f9), production code untouched -- pytest tests/test_godcmd_model_argument_count.py fails the first case with AssertionError: assert 'string' != 'string', i.e. the reply the user is handed really is the placeholder.
• GREEN after the change: 3 passed.
• Whole suite on the same tip, before and after: the FAILED set is identical (32 pre-existing failures, all environment-bound on Windows -- scheduler timezone, symlink pruning, the rg backend parity case, /proc/self/environ probes). passed goes 2259 -> 2262, exactly the three new cases.
• Repo's own lint gate (ruff check --select F --ignore F403,F405) on both changed files: All checks passed.

The new tests build the plugin against a tmp_path config carrying admin_users and stub conf() and Bridge, so no network is touched and no Bridge().reset_bot() is called; conf()["model"] is never written.