Measure CACHE HIT % based on total cache misses and the theoretical total number of build rules.
REASON: If we only look at cache_misses and processed rules we are strongly biasing the result
toward misses. Basically misses weight more than hits. One MISS will traverse all its dependency
subtree potentially generating misses for all internal Nodes; worst case generating N
cache_misses. One HIT will prevent any further traversal of dependency sub-tree nodes so it will
only count as one cache_hit even though it saved us from fetching N nodes.