Add  % of total for IO and Compute to `--analyze`

This is a bit nuanced because of parallelism / the io metrics being in wall time / using cpu time for compute metrics.  Its definitely possible there is data that can be used for this that i just didnt find an API for yet.