Right now we're sampling 100 documents from each collection for generating the initial data model of the user's schema. This ticket involves re-evaluating that number. It may be something we should set it based on the amount of collections, or let the users themselves set. Or we can increase that number if we find it best.

In the tech design: https://docs.google.com/document/d/1ScR8DgLMBpeyQvgSPQ5iLHXpkugekeAJ74sUIj2CjHs/edit?tab=t.0#bookmark=id.7xq1161xs44