Skip all-null inputs when merging min/max in Statistics::try_merge_iter #26748

Follow-up from #26310 (review):

skip all-null inputs when merging min/max in Statistics::try_merge_iter. Until that lands, the FileGroup doc (file_groups.rs:374) saying min/max are the partition bounds across the group is wrong for a group that contains a null partition.

A null partition (e.g. d=__HIVE_DEFAULT_PARTITION__/) has Absent min/max, and Precision::min/max return Absent if either side is Absent. So a single null partition drops the bounds for the whole group and table, even though every other file has exact bounds.