Support for order sensitive NTH_VALUE aggregation, make reverse ARRAY_AGG more efficient

[mustafasrepo]

Which issue does this PR close?

Closes #.

Rationale for this change

Currently, we have FIRST_VALUE, LAST_VALUE, ARRAY_AGG order sensitive aggregations functions. Similar to the window counter part one may need to index entries other than first and last. With NTH_VALUE aggregation function one can choose arbitrary indices. With this PR one can write following query, to get second row in the group.

SELECT a, b, NTH_VALUE(c, 2)
FROM multiple_ordered_table
GROUP BY a, b
ORDER BY a, b;

As a workaround following query can be used to produce desired output of the query above.

SELECT a, b, ARRAY_AGG(c)[2]
FROM multiple_ordered_table
GROUP BY a, b
ORDER BY a, b;

However, with dedicated aggregator, first query will use less memory.

What changes are included in this PR?

This PR adds NTH_VALUE aggregator support.

Are these changes tested?

Yes,

Are there any user-facing changes?

[ozankabak]

I reviewed this PR and it looks good except one issue about clones during merge_batch operations. There seems to be an inefficiency there which is not introduced by this PR, but I discovered it during review. I will consult with @mustafasrepo tomorrow to discuss possible fixes.

[ozankabak]

I reviewed this PR thoroughly again and the inefficiency I found in my first review is resolved -- we both have faster ARRAY_AGGs now, and also have an efficient NTH_VALUE.