While MongoDB has no schema, there are indexes in collections.  mongodump should allow dumping just metadata and no data, so that we can easily recreate a production environment's indexes into a dev cluster without dumping real prod data.

This can be achieved today but providing a query that matches no documents, e.g. `mongodump -q {"_id" : -1 }` but a specific documented parameter such as `-nodata` or `-metadata` would be more user-friendly.