When capturing a large number of measurements in a single benchmark, the nxn matrix becomes too large to grok.

I see #47 was opened to consider ways to filter the table, but for our current use cases of multiple-measurements, we really just want to compare measurement X's to other measurement X's and measurement Y's to other measurement Y's (can be separate tables).