Removes the assertion regarding the precompute factor needing to be a power of 2. There is no such requirement and it works just fine for other values too.
Fixes the average bucket size for the large buckets threshold - it depends on the precompute factor.
This PR fixes 2 things: