Planner: Small NDV results in too many histogram buckets.

## Enhancement
The concept behind histograms for optimizer statistics is to represent a range of values - it's cardinality, and number of duplicates per value. Having a very small number of values per bucket results in too many buckets - which drives up the number of statistics we manage.

There are exceptions - where capturing the ranges such that gaps appear between buckets vs within bucket could result in the need for more buckets - but sampling can distort this and mislead the optimizer.