Add ComputeLoss() class

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refactoring of loss computation to enhance modularity and scalability in YOLOv5.

📊 Key Changes

• Removed the global compute_loss function and replaced it with a ComputeLoss class.
• The ComputeLoss class now takes care of calculating the loss and is instantiated in train.py.
• Adjusted calls to compute loss in both train.py and test.py to use the new class.

🎯 Purpose & Impact

• 🛠️ Modularity: Encapsulating the loss function in a class improves code organization, making it easier to manage and understand.
• 📈 Scalability: The change prepares the codebase for potentially more complex loss computations and further development.
• 🚀 User Experience: Users now have a clear interface for loss computation, which may aid in custom modifications or extensions.