[REQUEST] Allow loading of YOLOv8 models from file instead of via YOLOWorld

[jnc-nj]

Judging by the issues in this repo, the Yoloworld ESAM node should probably allow loading of yolo models from disk rather than via roboflow. There's really no reason to use a third party service when the relevant models are freely available on hugginface.co.

Should probably also add a warning on the readme to stop peeps from installing both onnxruntime and onnxruntime-gpu (which somehow isn't blocked while installing inference).

[NeoLoger]

I was just looking into how I can load custom YOLOv8 model. +1 on this request this tool can become very powerful if you can load custom modules you trained yourself.