Cache pybids database for dataset to speed up loading large datasets (i.e. narratives)

[adelavega]

For very large datasets, constructing a database is time consuming.

It would be ideal to be able to pre-construct a dataset level cached pybids DB.

However, in NS, we need a database that also includes the specific events that are indexed for a particular analysis. Thus, we can't currently simply cache the dataset db, and pass it to fitlins, as it would not include the events.

In order to do so, we would need to re-create the pybids layout in Neuroscout-CLI for each dataset, save it in a standardized location (i.e. in the dataset itself) and then add the events for that specific analysis, and pass the cached db location to fitlins