Perform additional work when updating the database to speed-up `DBExtract` generation

[paulmueller]

This could include:

• fetching all collections that a user can add datasets to (saves at least one API call to group_list_authz)
• creating an index for all collections and circles: Might require separate calls on update, since adding datasets to a collection does not update the dataset metadata and thus does not show up in simple dataset search with time_since. Fetching a list of all datasets for all collections might take some time.