As a follow up to SERVER-109942, we can make some improvements on the basic algorithm. This ticket should cover

1. Handling multiple hashPrefixes (returning multiple bucket ranges)
2. Bucket not just based on one additional byte, but dynamically based on the amount of hashPrefixes provided (calculation in design) to allow the response to still fit within 16MB
3. Any further optimization (not writing out empty hash buckets, do not perform normal validation checks during drill down, etc)

Existing tests under SERVER-109942 should be extended to cover multiple hash prefixes (or new tests if easier)