Spec deviation for Zarr scalar strings

[Bisaloo]

Description

Spec says:

In zarr, scalar strings MUST be stored as a fixed length unicode dtype

But in #455, both array strings and scalar strings were updated to use VLen-UTF8

[Bisaloo]

It would be worth double checking what anndata python is doing to determine whether:

• this needs to be updated in anndataR
• the spec is inaccurate and doesn't faithfully reflect what anndata python is doing

FWIW, both example_v2.zarr.zip and example_v3.zarr.zip in this repo use VLen-UTF8 for scalar strings.

[Artur-man]

@ilan-gold would be the best to answer this, and also sharing the script that generates both zarr stores:
https://github.com/scverse/anndataR/blob/devel/inst/scripts/example_files.py