pub fn scale(m: u64, e: i32) -> f64
m * 10^e or m / 10^-e (m < 2^63), rounded once to double [f64_parse .fp_calc]