This allows for testing floating-point audio encoders across different
platforms where exact comparisons are unreliable due to float rounding
differences.