fn align_parts(parts: &[BulkPart]) -> Result<(SchemaRef, Vec<DfRecordBatch>)>
Align the JSON columns in BulkParts, to unified Arrow arrays. So that we can compute (concat, sort, etc.) on them.