Sort indexes in an order that is likely to be a good index merge intersection order. After running this function, [start, ..., end-1] is ordered according to this strategy: 1) Minimize the number of indexes that must be used in the intersection. I.e., the index covering most fields not already covered by other indexes earlier in the sort order is picked first. 2) When mu
source not stored for this graph (policy: none)
no test coverage detected