fix: correct duplicate field error conversion for case-insensitive mode

Filter DataFusion's valid fields list to only case-insensitive matches
instead of passing all schema fields. This fixes the SPARK-25207 test
in ParquetFilterSuite which expects matched fields [B, b] not [A, B, b].

Also update the Spark test diff to accept both uppercase and lowercase
field names since DataFusion doesn't have access to the original table
schema field name.

Author: andygrove