Problem/Motivation
Per #3500683: Allow access policies to opt out of caching, performance tests currently throw all cache get/set/delete operations in one single count per operation.
However, different bins have very different performance implications, specifically fast chained bins are way faster. Issues like #3500687: Use bootstrap cache bin in ThemeExtensionList that move caches from regular to fast bins are not "seen" by performance tests at all.
Additionally, just having the count is a bit of a blackbox still. It's not always clear what/why there is a change. Maybe we should just store all cids as well so to simplify debugging and allow to compare/assert more specifically if so desired?
