This method compares floats using combined absolute and relative tolerances, instead of the "number of decimal digits" that TestCase.assertAlmostEqual uses.