Option to have relative allowed error in numbers but with also allowing a small absolute error

[dqnykamp]

When one generates answers through a numerical algorithm, it's common to have 0 become a small number on the order of 10^-15. In these cases, one would typically want any small number of that order to be considered correct, i.e. to have an absolute allowed error of around 10^-14.

Right now, we have a boolean "allowedErrorIsAbsolute", but in this case you want a combination of both relative error (for answers that don't end up being essentially zero) and a small absolute error (for the above case). Do we change how the allowed error attributes work so that you have to supply both allowed absolute error and allowed relative error? What should the defaults be?