We need to estimate the number of distinct values of an attribute frominside optimizer, for an extension we are working on.Should we use stacommonfrac from pg_statistic, or attdisbursion, or whatelse?