Napari plugin reader for videos

[niksirbi]

We would like to have the ability to load single frames from a video (minimally just loading an image) into napari as an Image layer.
The development of this feature will be facilitated by #38 and it will in turn facilitate #50.

[niksirbi]

Simple image frames can be just drag-and-dropped into napari and they will automatically open as images. What we want instead is a way to load videos. This is possible via the napari-video plugin, which is essentially a reader plugin for videos.

That said, it would be useful to add a visible place on the movement widget to load the video together with the predicted poses.

[sfmig]

When implementing this, check resources linked in #105

[niksirbi]

While working on #253, I ran into issues with napari-video /videoreader, see postpop/videoreader#6

OpenCV also seems to lead to inaccurate seeking (which will be important for labelling), see janclemenslab/napari-video#3

Given the above, might be worth checking out the newer napari-pyav plugin, which uses av as a backend, instead of opencv.

Another alternative would be relying on sleap-io to handle the video-loading, which currently uses an ffmpeg backend (via imageio-ffmpeg. The advantange is that we already depened on sleap-io so no new dependencies need be added. But in that case we would have to write our own napari plugin reader for videos - sth similar to napari-video and napari-pyav but using sleap-io.