Benchmark scoring function should be swappable

[rogthefrog]

For benchmark repeatability (e.g. "rerun this benchmark using the scoring used for v1) and evaluation (e.g. "compare scoring from v1 to this new scoring logic"), we want the benchmark scoring function from v1 to live forever AND the ability to plug in new scoring functions.