The lastpoint optimization and rewrite of group as transform on the first document and common sort key generation could have shared the great deal of code and yet both have limitations of its own.

The lastpoint optimization support only 1 accumulator and on the other hand, rewrite of group as transform on the first document does not support $topN/$bottomN with n=1.

If we define the shared code between two optimization, we should be able to overcome both limitations.

Also, if we are able to give the higher priority to the lastpoint optimization & rewrite of group as transform on the first document than to common sort key generation among multiple $top/$bottom accumulators while working on this, we potentially avoid BF-33077 regressions.