Persist provenance graph outside notebook

[sverhoeven]

Now the provenance graph is stored in the metadata field of the notebook file.

This is not nice because:

• Notebook should be without history
• Provenance graph gets big fairly quickly, storing it as json does not scale.

In https://github.com/activityhistory/nbcomet they use a sqlite file for each notebook in ~/.jupyter/nbcomet, we could do something similar.

[thinkh]

We could try turtleDB for client-side database and synchronization with the server. However, the backend part tortoiseDB (via a second package) is based on MongoDB and npm. So we probably have to write a server-side adapter for the SQLite storage.

[thinkh]

Other ideas:

• Research how jupyter checkpoints work and when a checkpoint is created?
• Alternative: Create subdirectory .ipynb_provenance and store provenance data as JSON files
• Evaluate if a local git repo for provenance data could be an option?