Revert C3() change

[glenn-jocher]

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Updating neural network architecture components in YOLOv5.

📊 Key Changes

• Removed CrossConv import from models/common.py.
• Switched from using CrossConv to Bottleneck in the C3 class definition.

🎯 Purpose & Impact

• The update emphasizes simplifying the CSP bottleneck layer by replacing a potentially experimental or less efficient CrossConv block with a more standard Bottleneck block.
• This change could improve the maintainability of the code, potentially enhance model performance, and ensure more stable behavior.
• Users might experience benefits such as faster inference, better accuracy, or reduced complexity when working with the YOLOv5 architecture.