Currently the compaction window summary displays the total bytes, over all input files, going into each compaction. To better observe the quality of the picked compaction, we ought to split the in(B) column into in_top(B) and in_bottom(B). Once the we land multi level compaction (tracked here) , we should add an in_extra(B) column as well. We should bikeshed the column names to our hearts content.

Jira issue: PEBBLE-55