[FEA] Simplify null pre-processing in sort-merge join

**Is your feature request related to a problem? Please describe.**
The lexicographic row comparator assumes null entries to be equivalent regardless of how nulls are to be compared in the join operation, which results in incorrect join output when `null_equality::UNEQUAL` is set. To circumvent this issue, the current sort-merge join implementation filters the input tables to remove all rows with null entries at the root level - and children level in case of lists - which can have a large memory footprint if the number of null rows is small.

**Describe the solution you'd like**
Support null comparison in lexicographic row operators based on the null equality set. A significant chunk of the pre-processing logic for sort-merge join can then be removed.