Enhancement - Add `config.yaml` for hyperparameters & settings

[Codex-Crusader]

Summary

Move hardcoded hyperparameters and runtime settings into a config.yaml (or config/default.yaml) and add a loader to read configuration at startup.

Motivation

A config file makes experiments reproducible and tuning easier without editing source files.

Acceptance criteria / Definition of Done

• Add configs/default.yaml with clearly documented sections (training, mcts, model, selfplay, io paths).
• Implement a small config loader (e.g., config.py) that:
  • Loads defaults from configs/default.yaml.
  • Allows overriding via CLI flags or environment variables.
• Update scripts (run_train.sh, run_selfplay.sh, run_play.sh) to accept or point to the config file.
• Document usage in README and provide one example config file committed to the repo.
• Add a unit test that loads the config and verifies required keys exist.

Implementation notes

• Use PyYAML (yaml.safe_load) or toml/json if preferred. Keep dependency footprint small.
• Ensure backward compatibility — if config missing, fallback to current defaults.

Complexity

Low to Medium — great starter task for contributors.