More uniform s-bucket sizes

[dariolina]

Right now, s-buckets of each record have a tail of "unlucky" ones that have way fewer chunks in them (see relevant forum discussion).
It's better to have them more uniformly sized, and a proposed solution is to modify quality lookup during plotting here to iterate backward. This should eliminate the tail, but central buckets may still be slightly fuller.

[nazar-pc]

This makes implementation and explanation significantly more complex with no clear benefit except nicely looking distribution of s-bucket sizes. We decided to not implement this.