Experimental Inverted Indexes Introduce Simplified Row ID Storage Heuristic

Experimental inverted indexes do not store tokens with too many matches (i.e. row ids in the posting list). This saves space and avoids ineffective index lookups when sequential scans would be equally fast or faster. The previous heuristics (density parameter passed to the index definition) that controlled when tokens would not be stored was too confusing for users. A much simpler heuristics based on parameter max_rows_per_postings_list (default: 64k) is introduced which directly controls the maximum allowed number of row ids in a postings list. #55616 (Harry Lee).