Apply cardinality thresholds before using indexes

`AndDocIdSet` evaluates sorted indexes, then bitmap indexes, and then scans, on the basis that these operators have increasing cost, but after any one of these stages it may be faster to skip indexes and go straight to scanning. This can be determined by checking the cardinality of the matching documents bitmap exceeds an empirically determined threshold before using an index.

Start by locating the AndDocIdSet entry point and trace how it evaluates sorted indexes, bitmap indexes, and scans. Determine where matching-document cardinality can be checked and how an empirically chosen threshold should affect index use; done means high-cardinality matches can bypass later index stages in favor of scanning.

Written by the indexing model from the issue text.