ENH: Add object detection model family, models, and datasets

[NickleDave]

• [ ] Add prep step for object detection datasets
• [ ] Add objection detection model family
• [ ] Add Deepsqueak model
  • [ ] Add Faster-RCNN that uses object detection model family
  • [ ] Add whichever version of YOLO they are currently using that uses object detection model family
• [ ] Should we also adapt voxaboxen? IIUC they only predict start/stop time but they are using OD models (something I keep thinking about doing, just to prove a point :thinking: :smiling_imp: )
• [ ] Add related datasets that I think actually have bbox annotations, e.g. MUPET, MouseTube