We should define an API that allows loading subdocuments into PyMongoArrow or update the documentation to explicit prohibit users from being able to do this.
Open questions:
- How will users define nested fields in the schema?
- Should PyMongoArrow's APIs allow users to flatten their documents?