Add support for batched inference, Ultralytics TRT inference, YOLOv6

[aliencaocao]

• Add support for batched inference on slices
• Add support for TensorRT for Ultralytics-based/originally YOLOv8 models (exported using Ultralytics)
• Add support for YOLOv6 (requires yolov6 repo on sys.path)

This PR is not meant to be merged but rather act as a reference for future more robust integration of these changes, as some of them are hacky patches and makes assumptions about the use case.

However, feel free to cherry pick changes from here and merge them, especially the batched inference part