New smart_resume()

[glenn-jocher]

Allow for commonality between tasks.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refined resume training functionality in YOLOv5's train.py.

📊 Key Changes

• Introduced smart_resume() function to clean up resume logic in training script.
• Removed repetitive code blocks by centralizing resume operations.
• Ensured optimizer and EMA (Exponential Moving Average) states are loaded correctly when resuming.
• Added clear logging about the remaining epochs when resuming training.
• Cleaned up label processing and moved the number of batches calculation (nb) to a more relevant code section.

🎯 Purpose & Impact

• 🔍 Streamlining: Simplifies the codebase for readability and maintainability.
• 🚀 Efficiency: Reduces errors and inefficiencies in the process of resuming training.
• 📈 Transparency: Provides better user feedback on training progress upon resume.
• 👤 User Experience: Enhances the clarity for users when they resume partially completed training sessions.