Add git info to training checkpoints

[glenn-jocher]

Assist with reproducibility, partially resolve #9643

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhanced YOLOv5 training logs with git metadata for better version tracking.

📊 Key Changes

• Added gitpython Module: A new dependency was added to the requirements.txt file to enable git-related operations within the code.
• Introduced GIT Variable: A GIT variable was created in utils/general.py to store the repository's remote URL, branch, and latest commit hash.
• Updated Training Artifacts: Training outputs, such as model checkpoints, now include git information to couple the code version with trained models.

🎯 Purpose & Impact

• Improved Reproducibility: Including git metadata ensures that training outcomes can be traced back to the exact code version, facilitating error tracking and experiment reproducibility.
• Enhanced Debugging: Developers and users can more easily relate model performance to specific commits, aiding in debugging and model improvement endeavors.
• Streamlined Workflow: This addition simplifies identifying the codebase status when sharing or reviewing models, leading to more streamlined collaboration among developers and researchers.