When using AutoLoader in a streaming application to ingest new files from an S3 location, it infers the schema based on the first 50 GB of data or the first 1000 files, whichever is less. Which configuration should be adjusted to change the default number of files used for schema inference to 500 for all future queries?