feat: improve accuracy of configuration forecast

The previous way was a bit silly, and just checked binary caches for all
of the requisites of the closure of a configuration -- meaning it was
checking for *derivations* of the build dependencies in binary caches,
rather than the out paths

Only the out paths of the build time closure are queried now, resulting
in a much more accurate (to what you would expect) forecast, as well as
way less queries made to a given cache

0 files changed, 0 insertions, 0 deletions