Add val.py no label warning

[glenn-jocher]

Help resolve confusion around zero-metrics val.py results when no labels are found in #8753

🛠️ PR Summary

_{Made with ❤️ by Ultralytics Actions}

🌟 Summary

Improved result handling for edge cases in validation script.

📊 Key Changes

• Added a check to handle cases where there are no targets (nt.sum() == 0) during validation.

🎯 Purpose & Impact

• 🛠️ Purpose: The change prevents potential division by zero errors or misleading results when there are no ground truth objects (targets) to compare against.
• 💻 User Impact: Users will experience more robust performance when validating datasets that might have images with no annotations. This can be particularly helpful when working with sparsely labeled datasets or when encountering edge cases during model validation.