Add local Qwen3 agent scaffolding

[nschlaepfer]

Summary

• add placeholder llama.cpp submodule and build scripts
• switch project to pyproject.toml
• implement basic agent package with tool wrappers and config
• add README instructions and CI workflow

Testing

• python -m pytest -q (fails: ModuleNotFoundError: No module named 'pytest')
• python -c "from agent.main import chat; print(chat('2+2?'))" (fails: No module named 'llama_cpp')