separate the raw stats from the index one - different libraries than index can use the raw.ml (for instance cactus), and this allows to have the IO stats even if index is not used;
different stats for merge and the rest of the operations - now stats are reported by file name and so we can report IO stats only due to the merge
This PR has two objectives:
raw.ml
(for instance cactus), and this allows to have the IO stats even ifindex
is not used;merge
and the rest of the operations - now stats are reported by file name and so we can report IO stats only due to the merge