Filter multiple columns from TopK using Lexicographical ordering

[Dandandan]

Is your feature request related to a problem or challenge?

In the PR #15697 we added support for filtering input values early on to speed up TopK execution.

The current code uses only the first value as filter (col[0] < threshold[0] for single columns and col[0] <= threshold[0] for multiple columns), but we could extend the expression to filter out multiple columns (either all of the columns or a prefix).

Describe the solution you'd like

We can write the filter that respects lexical ordering as:

col[0] < threshold[0] || (col[0] = threshold[0] && col[1] < threshold[1] || (...))

Describe alternatives you've considered

No response

Additional context

No response

[Standing-Man]

take

[Dandandan]

This waits on

Https://github.com/apache/datafusion/pull/15697 to be finalized