Fix fp16 (--half) support for TritonRemoteModel model type

[fabito]

Fixes: #10786

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Enhancement of mixed-precision inference support.

📊 Key Changes

• Allow FP16 (half-precision) inference on Triton backends.

🎯 Purpose & Impact

• 🚀 Purpose: To enable more efficient memory usage and potentially faster inference times when using the Triton Inference Server by supporting half-precision (FP16).
• ⏱ Impact: Users deploying models with Triton can take advantage of reduced memory footprint and increased performance due to FP16 computation. This can be particularly beneficial for GPUs with Tensor Cores that are optimized for FP16 operations.

[github-actions]

👋 Hello @fabito, thank you for submitting a YOLOv5 🚀 PR! To allow your work to be integrated as seamlessly as possible, we advise you to:

• ✅ Verify your PR is up-to-date with ultralytics/yolov5 master branch. If your PR is behind you can update your code by clicking the 'Update branch' button or by running git pull and git merge master locally.
• ✅ Verify all YOLOv5 Continuous Integration (CI) checks are passing.
• ✅ Reduce changes to the absolute minimum required for your bug fix or feature addition. "It is not daily increase but daily decrease, hack away the unessential. The closer to the source, the less wastage there is." — Bruce Lee