Add Paddle exports to benchmarks

[glenn-jocher]

Signed-off-by: Glenn Jocher glenn.jocher@ultralytics.com

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to YOLOv5 inference support and data processing.

📊 Key Changes

• Removed PaddlePaddle's output handle redundancies for efficiency.
• Enabled inference on all export formats except for Edge TPU and TF.js.
• Improved data type consistency and inference support across different formats.
• Simplified the way PaddlePaddle outputs are handled during forward passes.
• Updated mask data types from np.bool to Python's built-in bool for better compatibility.

🎯 Purpose & Impact

• 🚀 Enhances the model's compatibility and performance with various export formats, except for Edge TPU and TF.js, expanding the range of supported devices and platforms.
• 💡 Simplifies PaddlePaddle's code for output processing, potentially improving maintainability and reducing the likelihood of future bugs.
• 🧠 Changes to data types promote consistency across the codebase, likely reducing errors and ensuring smoother data handling processes.
• 🖼 Updates to image and mask plotting improve reliability and follow best practices in type usage, potentially affecting all visual outputs or segmented images, enhancing visualization quality.