prioritize low selectivity indexes in the greedy search procedure of Selectivity()

[eurekaka]

Development Task

Consider this case:

create table t(a int, b int, c int, d int, key ab(a,b), key cd(c,d), key bc(b,c));
explain select * from t where a = 1 and b = 1 and c = 1 and d = 1;

When computing the total selectivity of this filter, current greedy algorithm would choose index ab and cd to cover the filters, and multiply these 2 selectivities computed.

However, if column b and c have some explicit or implicit functional dependency, specifically, there is no record satisfying b = 1 and c = 1 at all, the final selectivity computed would be too high. Instead, if we can put index bc into the covering index set, then the final selectivity would be the correct 0.