Add F1 score metric to detection metrics

[kayampady]

This PR adds F1 score as an evaluation metric for detection tasks.

F1 score provides a balanced measure of precision and recall, complementing the existing metrics in the detection pipeline.

The implementation integrates with the current metric computation framework and follows the existing design patterns.

[kayampady]

The failing tests appear unrelated to the F1 metric changes in this PR.

All detection and segmentation metric tests pass successfully.

The failure is coming from tests/test_lidar.py with an Open3D-related TypeError in CI:
isinstance() arg 2 must be a type

This seems to be an environment/dependency issue with Open3D.

Related issue: #556

Happy to help debug or take this up in a separate PR if needed.

[kayampady]

Closing this PR as it duplicates the earlier F1 implementation.

Will reopen PR #478 and link it with the newly created issue #556 as suggested.