Comment 2 for bug 1505286

So, thinking more on how the limit should be expressed, I think it should be per facet type. For example, consider a search results set that has 10,000 subject facets having two more more records, and 1 author facet on just one record. Picking the top 1,000 facets among the total set loses the author facet, whereas picking the top 100 (or whatever) facets for each type doesn't drop the author facet.