fbm can be a lot faster

[fmrl-island]

Hi,
I noticed that in fbm.rs (2d, 3d, 4d), the signal is multiplied by self.persistence.powi(x as i32); each time through the loop. My code runs in only 45% of the time if I change that to instead by a local variable that is updated each octave. I think it is still correct. Thanks!

   fn get(&self, point: [f64; 2]) -> f64 {
        ...

        let mut temp = 1.0;

        for x in 0..self.octaves {
             ...

            // Scale the amplitude appropriately for this frequency.
            signal *= temp;
            temp *= self.persistence;
            ...
        }
         ...
    }
    ```