Prevent duplicate columns in the schema when partitions are present in parquet files

[rtyler]

In some scenarios Big Query can inline a partition column in output parquet files and some deduplication needs to happen on columns before the initial commit on the table gets created