Refactor new model.warmup() method

[glenn-jocher]

Adds warmup to TRT and ONNX models to avoid first batch timing artifacts. May resolve #5806

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refactoring model warmup process in Ultralytics YOLOv5.

📊 Key Changes

• Consolidated model warmup code into a single method warmup() within the Model class.
• Removed conditional warmup checks from detect.py and val.py and replaced them with calls to the new warmup() method.

🎯 Purpose & Impact

• Code Maintainability: Centralizing warmup logic makes the codebase cleaner and easier to maintain.
• Consistency: Unified warmup approach ensures consistent behavior across different scripts within the repository.
• Performance: The warmup is reserved for GPU use, potentially avoiding unnecessary resource usage on CPU devices.
• End-User Impact: Streamlines internal code with no direct feature or performance changes for the end-users but sets the stage for a more robust and maintainable system.