Store a compressed bitmap alongside the index pages

When we match a page in the btree search we then delegate to the sub-index to search that page. This works well for high cardinality data and equality / inequality searches. However, it can be inefficient when the data is low cardinality since we have to search a very large page (or many small pages)

Luckily, when we search the btree, we can know if a page is entirely included in the result. In other words, if a page has min: 5, max: 10 and the query is value < 20 then we don't need to search the page because we know we want all of the rows in the page. If we had stored a compressed bitmap of all the row ids in the page then all we would need to do would be to load that.

Start by tracing the btree search and the delegated sub-index search described in the issue, then identify how index pages and their row ids are stored and loaded. Done means storing a compressed bitmap alongside each index page and loading it when the btree shows that the page is entirely included in the query result.