buildScan for a BaseRelation that can eliminate unneeded columns before producing an RDD.
buildScan for a BaseRelation that can eliminate unneeded columns before producing an RDD. It uses csv data source to read the gff file. In addition to base columns, it parses "attributes" column to populate the flattened attribute columns in the schema in correct data type. If schema contains "attributes" columns the original attributes column will be included as well.