Cost per entry for Lucene index of type v1 should be higher than that of v2

Currently default cost per entry for Lucene index of type

1. v1 - which uses query time aggregation
2. v2 - which uses index time aggregation

Are same. However given that query time aggregation would require more effort it should result in a higher cost per entry.

This fact impacts the result in cases like OAK-2081 (see last few comments) where with usage of limits both index are currently considered equals