Update val.py speed and study tasks

[glenn-jocher]

Accepts all arguments now by default resolving #5600

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Refinement of validation benchmarks and study tasks in YOLOv5.

📊 Key Changes

• Consolidated benchmarking and study tasks under a single conditional structure.
• Enabled FP16 (opt.half = True) by default for the 'speed' task to enhance inference speed.
• Set default confidence (opt.conf_thres) and intersection over union (opt.iou_thres) thresholds for the 'speed' task.
• Streamlined the loops for processing multiple weights by directly iterating over them.
• Simplified data output code to write results into files for the 'study' task.

🎯 Purpose & Impact

• 🚀 Simplifies the codebase and improves maintainability by reducing complexity in val.py.
• ⏱ Enhances the performance (speed) of benchmarking tests by defaulting to FP16 precision, making benchmark results more reflective of real-world fast deployment scenarios.
• ✅ Users can now more easily perform 'speed' and 'study' benchmarks with consistent and standardized settings.
• 📈 Improves the ease and consistency of capturing and reviewing results from model study benchmarks, aiding in research and development efforts.