Create a testing dandiset with a lot of metadata / metadata variation

@satra wrote:
> for the future, since it is hard to compare DLP completeness before and after on staging, we should craft a staging dandiset that has as much metadata/metadata variation as possible.

(original discussion in https://github.com/dandi/dandiarchive/issues/1012#issue-1092810649)

Start with the original discussion in dandi/dandiarchive issue 1012 and review how staging DLP completeness is currently compared. Define the metadata and variation the testing dandiset must cover, then create it in staging and verify that it exercises the intended completeness checks.

Written by the indexing model from the issue text.