Module rrf 

Reciprocal Rank Fusion: score(i) = sum_b 1 / (k0 + rank_b(i)).

Rank is 1-based. Missing ranks do not contribute. No score calibration. k0 is typically 60.

Cormack, Clarke, Buettcher, Reciprocal Rank Fusion outperforms Condorcet and individual Rank Learning Methods, SIGIR 2009. doi:10.1145/1571941.1572114

Functions

rrf_merge

Merge ballots by Reciprocal Rank Fusion. Ties break by first-seen id order.