Insertion sort for consolidation

We very often "consolidate" collections of updates, triples (data, time, diff), so that there is at most one element with each data and time. This is currently done by sorting such a collection by (data, time) and then swinging through the result.

Most sorting algorithms use insertion sort on small sets, either in the recursive base case for quicksort or to form sorted runs for mergesort. We could perform a similar action when adding elements to a "compactable list", with the improvement that if we find a matching element it can be compacted in place. Ideally this would be no worse than sorting the data (although it probably does work that subsequent sorting can't exploit perfectly), and often much better (when adjacent elements collapse, or when the set of data and times is small).

More generally, we should be mindful that consolidation is not always cheap (we must allocate the backing memory, and then perform super-linear sorting), and a solid implementation of compaction should have pervasive performance wins.

The issue names no files, tests, or entry points to begin from. Investigate the existing compactable-list and consolidation implementation, then establish benchmarks and tests showing insertion-time compaction is correct and improves consolidation performance before defining the implementation scope.