Video support: Adding models for video and object detection track

[JonasWurst]

Adding the models for video and video object detections. Here, the format like in youtube-vis is selected.

points that will be addressed in a follow up:

• adding frame list to the video class to allow mapping back to src data
• adding confidence values to the Object detection

Open question:

• we could opt for a more explicit linking between frames and their annotations. At the moment we have independent lists which can be error-prone. Suggestions are welcome

Note:
This PR was inspired by #27 by @fardinayar