Consider Extending Dask Dataframes to provide missing key features.

[alxmrs]

By subclassing ddf, we could provide a few key features to make Qarray better:

• Centralized Metadata, like df shape / size (#18)
• MultiIndex indexes, like pandas. These would be critical for efficient search along coordinate axes.

[alxmrs]

https://docs.dask.org/en/latest/dataframe-extend.html

[alxmrs]

Consider adopting Julia Signell's MultiIndex in for Dask Dataframes: https://github.com/dask/dask/pull/8153