Add Backgammon game, pure MCTS agent, and pygame UI

[ManuelFay]

Motivation

• Provide a Backgammon implementation analogous to the existing Connect4 codebase to enable MCTS-based play.
• Offer a pure-MCTS agent baseline to exercise game rules and move-generation before adding neural MCTS.
• Add a simple pygame-based UI and gameplay loop so the new game can be played and visually inspected.
• Expose the Backgammon agent from the package so it can be used like other agents.

Description

• Add BackgammonBoard implementing rules, move generation, dice handling, and legal-action sequencing in alphaconnect4/interfaces/backgammon_board.py.
• Add BackgammonTree MCTS node wrapper in alphaconnect4/interfaces/backgammon_mcts_interface.py to adapt board states to the existing MCTS engine and provide find_children, find_random_child, and reward.
• Add BackgammonMCTSAgent in alphaconnect4/agents/backgammon_mcts_agent.py that uses MCTS for move selection and reports ai_confidence.
• Add UI constants, a pygame visual engine (gameplay/backgammon_visual_engine.py), a gameplay loop (gameplay/backgammon_game.py), and a runnable entrypoint run_backgammon.py, and export the agent from alphaconnect4/agents/__init__.py.

Testing

• No automated tests were executed as part of this change.
• The new files were exercised interactively during development (visual/manual checks) but not via CI-run unit tests.
• No test failures were reported because no test suite was run.
• Further automated tests (unit tests for move generation and MCTS behavior) are recommended before merging into mainline.

---

Codex Task