Return page numbers on main search results #442

Consider WRAL's UNC doc searcher. When you perform a text search, it hits the main /api/search.json, collects all the resulting document IDs, and then ❌ performs an individual search against each resulting document with /documents/search.json. So, in that case, searching for something like "UNC" results in 1 ajax request for the initial search, then 1000 ajax requests to get the page numbers for where the text appears in each document.

This could all be collapsed if /api/search.json would go ahead and return what page numbers text appeared on (like /documents/search.json does) when text was one of the search facets.