Add high-level and technical documentation about DataLad datasets

[kabilar]

It would be helpful to add high-level documentation that describes the motivation and use cases for DataLad datasets that mirror the Dandisets, as well as technical documentation that describes how to use these datasets.

Relevant repositories:

1. https://github.com/dandi/dandisets
2. https://github.com/dandisets
3. https://github.com/dandi/backups2datalad

[yarikoptic]

Very good point! I think there are two components -- one is user facing and another one "development/infrastructure".
If I don't get to it -- you are most welcome to take a stab at it.

Some other points to mention:

• there is also https://github.com/dandizarrs/ -- collection of individual per zarr directory DataLad datasets
• https://datasets.datalad.org/?dir=/dandi contains "once in a while" updated clones of the dandisets
• https://registry.datalad.org/ based on auto-discovered registry of datalad datasets at https://github.com/datalad/datalad-usage-dashboard/ provides extra functionality to search metadata within dandisets and also for reuses of dandi datasets (when forked or linked as subdatasets).

While describing backups2datalad IMHO worth mentioning

• Initial motivation was to establish backup of DANDI
• ATM it provides the best mechanism to access prior versions of draft versions of dandisets.
• ATM all assets (and partially zarrs) are backed up on dropbox using https://github.com/git-annex-remote-rclone/git-annex-remote-rclone/ (which we now maintain)