If the `m` value in the CPR calculation works out to be negative the use
of `%` results in an incorrect (negative) value.

This commit uses `rem_euclid` instead of `%` to get the expected value
when calculating longitude (at least for the test input used and within
a few decimal places).

I'm not entirely sure that `rem_euclid` is the correct replacement but
from the little bit of reading I've done to refresh myself on modulus
and related math I think it is suitable (plus the existing test suite
still passes in addtion to the new test for this case).

Example Rust code to show differences:

    #![allow(unused)]
    fn main() {
        let ni = 47.0;
        let m: f64 = -28.0;
        println!("ni = {}", ni);
        println!("m = {}", m);
        println!("m % ni = {}", (m % ni));
        println!("m.rem_euclid(ni) = {}", m.rem_euclid(ni));
    }

Which yields:

    ni = 47
    m = -28
    m % ni = -28
    m.rem_euclid(ni) = 19

Which is the same as Python `%`:

    In [1]: -28 % 47
    Out[1]: 19