video support

[csbrown]

We have a lot of video data with associated bounding-box and keypoint annotations that we would like to Croissantify. It seems that the Croissant spec does not currently support video. I am willing and able to put in some work toward extending the spec and implementing tools in the python library.

However, I'm new here and am not super sure where to start. Can someone provide some pointers on where I might start to delve to get a handle on what it would take to implement this?

Related #690 #371