Rank fallback reviewers by the count of files that the fallbacks apply to

if there are many path fallbacks that apply to a PR where 70% of the files have a single applicable fallback, contrary to user expectations, the likelihood that a reviewer from the single fallback is chosen is 1/(all reviewers from all fallbacks). one improvement could be to sort fallbacks by applicable file count and drop the lowest "scoring" ones

No file, test, or entry point is named. Start by locating fallback reviewer selection and how applicable files are counted, then clarify whether low-count fallbacks should be ranked lower or excluded. Done means reviewer selection reflects the number of files each fallback applies to, including the described 70% case.

Written by the indexing model from the issue text.