once we have built our actual final "raw-data" in the mapping
hash, we iterate all runners and sort their match lists.

this in particular prevents matches to get modeled in order of
appearance rather than relevance, which for finely relevance sorted
runners such as the services runner is very important to give
the most suitable results first.

• add some chrome sites to desktop
• search for chrome in kickoff
• results ordered as per services runner relevance with chrome itself at the top of the list