sys::scale-exponent was not working correctly and outputting
    numbers in the wrong range. Furthermore, using sys::scale-exponent
    for scaling the number to the correct range is flawed anyway,
    since it introduces rounding errors. Hence we replace
    sys::scale-exponent by the much simpler sys::exponent-in-base10
    function and fix the logic in format-exp-aux to scale using
    sys::flonum-to-string, which is rounding error free. Moreover,
    sys::flonum-to-string was buggy and not rounding numbers correctly
    when the 'd' parameter was given, which has also been fixed.
    Fixes #437.