compares two sets of values to see if they remain within epsilon of each other
using relative error calculation of val1-val2 divided by the average of val1 and val2 (to avoid dividing by 0
while maintaining similar range of values to regular relative error)