Add --int8 argument

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to model export functionality with INT8 quantization support for TensorFlow Lite. 🚀

📊 Key Changes

• Avoids adding ROOT to sys.path if it's already present.
• Renames the tfl_int8 parameter to int8 for clarity.
• Adds INT8 quantization support for CoreML and TensorFlow Lite model exports.

🎯 Purpose & Impact

• Prevents unnecessary duplication in sys.path, leading to cleaner code execution.
• The clearer parameter naming (int8) improves code readability and understanding.
• Enables users to export TensorFlow Lite models with INT8 quantization, potentially reducing model size and improving performance on INT8-capable hardware. This change expands the utility of YOLOv5 for mobile and edge devices with optimized inference capabilities. 👍💨