Adds RKNN wheels and checksums

[Burhan-Q]

Per discussion from YOLO meeting on August 12, 2024 with @glenn-jocher, adding wheel files for installing RKNN_toolkit2 for use with ultralytics.Exporter and ultralytics.AutoModel classes to support RKNN model conversion.

• RKNN_toolkit2 is specifically for x86 platforms (export on desktop)
  • wheel files were too large to upload to GitHub (~39 MB)
• RKNN_toolkit2-lite is for ARM platforms (inference and export)

Note

merging this branch may not be required at all, as long as URLs to files are static

FYI: @lakshanthad and @IvorZhu331

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Added RKNN Toolkit2 Lite and RKNN Toolkit2 resources for inference.

📊 Key Changes

• New Readme Files: Added documentation for RKNN Toolkit2 Lite and RKNN Toolkit2.
• Wheel Files: Included RKNN Toolkit2 Lite wheel files for multiple Python versions (cp38, cp39, cp310, cp311).
• MD5 Checksum Files: Added MD5 checksums to verify integrity of wheel files.
• Requirements Files: Listed dependencies for different RKNN Toolkit2 versions and Python versions.

🎯 Purpose & Impact

• Enhanced Documentation: Facilitates better understanding and usage of RKNN Toolkit2 Lite.
• Compatibility & Integrity: Provides necessary wheel files and their checksums, ensuring compatibility with various Python versions and integrity check mechanisms.
• Dependency Clarity: Clear requirements listed for easy setup, improving user experience during installation and reducing potential setup issues.