Load 3D pose data from Anipose

[niksirbi]

Is your feature request related to a problem? Please describe.
We currently only load 2D pose predictions from SLEAP, DeepLabCut and LightningPose. We want to expand our support to 3D as well, e.g. Anipose and/or DANNCE.

Describe the solution you'd like
Anipose seems like a good starting point as we feel it's widely used. We should get ourselves some sample output files to get working on this.

Additional context
Working on this will also reveal if/how our current data model is inadequate for representing 3D data.

[talmo]

FWIW:

https://github.com/talmolab/sleap-anipose

https://github.com/dattalab/keypoint-moseq/blob/a3c569381dc1ce8040dce1afd6ac7ddc945ce187/keypoint_moseq/io.py#L1027-L1065

Our data model implemented in sleap-io doesn't support 3D yet either, but ndx-pose does.

[niksirbi]

Thanks for the super-specific links. These make it extremely easy for us to also implement an anipose-loader.

[niksirbi]

Another 3D format would be that of JARVIS.

Following this manual we should be able to create example prediction files data3D.csv.

[vigji]

I have written a first draft of an anipose loader, can open a PR if this is not been addressed anywhere!

[niksirbi]

By all means do open a draft PR @vigji, none of us is working on it at the moment.
Feel free to just dump what you have in the first instance, after we take a look at it we can make plan on how to best integrate it into the existing API.