Split-up S3 batch export file according to file size limit

[tomasfarias]

Large files can be hard to manage, so our S3 batch export should support splitting-up a file into individual parts according to a max file size provided by the user.

Some questions are open like when exporting Parquet files we write the entire Arrow RecordBatch, which could be larger than the max file size, so the writer needs to handle this limit too. JSONL S3 batch exports should be easier to implement as they write rows one-by-one.