Major Description
Counting up facets should take advantage of concurrent hardware if the index has multiple segments.
At first I tried to add an optional ExecutorService argument to the existing single-threaded SortedSetDocValuesFacetCounts but it was difficult, so I just made a new class.
I also internally refactored SortedSetDocValuesFacetCounts to share code between its count and countAll private methods, and switched the FacetsCollector to use DocIdSetBuilder instead of always using a FixedBitSet.