[WIP] Video Annotation on CVAT-MONAI Label with Active Learning

[tangy5]

The draft PR shows the development progress, not ready for review yet.

The draft PR adds the video frame loader by cv2 and transforms. Added average strategy for video scoring. Select top K videos and push to CVAT server for annotation.

ToolTracking model is used for sample implementation.

Next step:

After done video annotation in CVAT, trigger video frame training and second round scoring for active learning.