[spark] Fix min and max query failure on metadata columns

[ArnavBalyan]

Purpose

• Min and max queries directly aggregating Paimon metadata columns fail.
• This is because agg pushdown is checking the physical schema for metadata columns. However, metadata columns are not in the physical schema, instead derived from split metadata as virtual columns.
• Skip pushdown for such columns and let Spark aggregate to avoid the failure.

Tests

• UT

[JingsongLi]

+1