Summary

• Fix a bug where flatten silently dropped data when the parent record had a conflicting column name that appeared after the column being flattened
• The conflict detection only checked already-inserted columns, missing conflicts with columns not yet processed
• Now collects all parent column names upfront into a HashSet before the loop, so conflict detection works regardless of column order
• Added regression tests for both the table and record cases

Before

> [[b, a]; [[[a]; [9]], 1]] | flatten -a b
╭───┬───╮
│ # │ a │
├───┼───┤
│ 0 │ 1 │
╰───┴───╯
# Inner value 9 is silently lost!

After

> [[b, a]; [[[a]; [9]], 1]] | flatten -a b
╭───┬─────┬───╮
│ # │ b_a │ a │
├───┼─────┼───┤
│ 0 │   9 │ 1 │
╰───┴─────┴───╯

Release notes summary

Fixed a bug where flatten would silently drop data when a nested record/table had column names conflicting with parent columns that appeared later in the record. Conflicting columns are now correctly renamed regardless of column order.

Closes #13271