Support annotating video-level attributes

[ehofesmann]

Currently, only frame-level labels are drawn on videos. https://github.com/voxel51/eta/blob/8e32f2a05f2754440b93a178ada743890158e50b/eta/core/annotations.py#L720

It would be useful (and necessary for FiftyOne Classification labels) to draw video-level attributes on videos as well.

Ex: Video-level video_labels to draw on each frame

{
    "attrs": {
        "attrs": [
            {
                "type": "eta.core.data.CategoricalAttribute",
                "name": "ground_truth",
                "value": "test"
            }
        ]
    }
}