"Binderize" repo

[chekos]

The original dataset could be broken down into ~10 pieces, uploaded to GitHub and in postBuild put back together.

This could help people prototype analyses the data further using python or R without having to download the data.

Ideas

• Could include some example notebooks with recreating published analyses on the dataset.
• Could also write a jupyter extension to deploy datasette on the mybinder instance. (How useful would this be though?)

[chekos]

Maybe add a "binder-stable" tag that mybinder would create the image against. metadata.json might be updated many times but the contents of the repository required for the data analysis shouldn't change that often.