Add --optimize argument

[glenn-jocher]

Fix for c++ runtime errors in #2973

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improvement in model export options with mobile optimization for TorchScript models.

📊 Key Changes

• Added a new command line argument --optimize which enables optimization for TorchScript models targeted at mobile devices.
• Modified the export logic to conditionally optimize TorchScript models using the new --optimize flag.
• Removed an unnecessary comment in the Core ML export section.

🎯 Purpose & Impact

• The new --optimize flag allows developers to export their models in a format better suited for mobile devices, potentially improving performance on such platforms.
• Conditional optimization ensures developers have the choice to either create standard TorchScript models or opt for mobile-optimized versions based on their deployment needs.
• By cleaning the codebase (removing the comment), the export script is simplified, making it easier to maintain and understand.

📱 For users deploying YOLOv5 models to mobile devices, this change could result in more efficient models, with potentially faster inference times and a smoother user experience.