Use faster loop iter (despite clippy's advice)

    $ cargo bench --bench="*" rust --  --baseline=main-2024-01-25
    direct (rust impl)/default
                            time:   [27.773 µs 27.806 µs 27.842 µs]
                            change: [-4.9255% -4.6540% -4.3868%] (p = 0.00 < 0.05)
                            Performance has improved.

    inverse (rust impl)/default
                            time:   [70.088 µs 70.139 µs 70.194 µs]
                            change: [-9.8981% -9.7426% -9.5881%] (p = 0.00 < 0.05)
                            Performance has improved.

src/geodesic.rs

@@ -185,20 +185,26 @@ impl Geodesic {
     pub fn _C3f(&self, eps: f64, c: &mut [f64]) {
         let mut mult = 1.0;
         let mut o = 0;
-        for (l, c_item) in c.iter_mut().enumerate().take(self.GEODESIC_ORDER).skip(1) {
-            let m = self.GEODESIC_ORDER - l - 1;
+        // Clippy wants us to turn this into `c.iter_mut().enumerate().take(geodesic_order + 1).skip(1)`
+        // but benching (rust-1.75) shows that it would be slower.
+        #[allow(clippy::needless_range_loop)]
+        for l in 1..GEODESIC_ORDER {
+            let m = GEODESIC_ORDER - l - 1;
             mult *= eps;
-            *c_item = mult * geomath::polyval(m, &self._C3x[o..], eps);
+            c[l] = mult * geomath::polyval(m, &self._C3x[o..], eps);
             o += m + 1;
         }
     }
 
     pub fn _C4f(&self, eps: f64, c: &mut [f64]) {
         let mut mult = 1.0;
         let mut o = 0;
-        for (l, c_item) in c.iter_mut().enumerate().take(self.GEODESIC_ORDER) {
-            let m = self.GEODESIC_ORDER - l - 1;
-            *c_item = mult * geomath::polyval(m, &self._C4x[o..], eps);
+        // Clippy wants us to turn this into `c.iter_mut().enumerate().take(geodesic_order + 1).skip(1)`
+        // but benching (rust-1.75) shows that it would be slower.
+        #[allow(clippy::needless_range_loop)]
+        for l in 0..GEODESIC_ORDER {
+            let m = GEODESIC_ORDER - l - 1;
+            c[l] = mult * geomath::polyval(m, &self._C4x[o..], eps);
             o += m + 1;
             mult *= eps;
         }