Yesterday a brief issue occurred where the server reported an incorrect
(too high) total number of results for a URL due to the search index
being out of date. Once the client reached the end of the actually
available results, it made queries to the search endpoint which returned
an empty result set but had a `total` figure implying that there should
be more pages. The client then ended up polling the server indefinitely
for more results.

This commit defensively makes the client stop fetching more result pages
if the current page is empty.