PyTorch Hub load directly when possible

[glenn-jocher]

This should reduce the verbosity and reduce time-to-load when loading official YOLOv5 PyTorch Hub models.

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancements to model loading and customizability in YOLOv5's hub interface.

📊 Key Changes

• Simplified loading of pretrained models with attempt_load function for standard use cases.
• Enhanced create function with additional autoshape and verbosity options.
• Improved error messages for outdated cache scenarios with explicit instruction to use force_reload=True.

🎯 Purpose & Impact

• The update streamlines the process of loading pretrained models, which could lead to simpler and cleaner user code.
• Adding the autoshape flag and verbose settings directly in the create function allows users more control over model initialization and logging.
• Clarified error messages aim to facilitate troubleshooting for users encountering cache issues.
• 📈 For users, these changes may lead to an improved experience when customizing and deploying YOLOv5 models, with potential time savings and reduced complexity in managing model configurations.