Fix array::IntoIter::fold to use the optimized Range::fold

It was using `Iterator::by_ref` in the implementation, which ended up pessimizing it enough that, for example, it didn't vectorize when we tried it in the <https://rust-lang.zulipchat.com/#narrow/stream/257879-project-portable-simd/topic/Reducing.20sum.20into.20wider.20types> conversation.

Demonstration that the codegen test doesn't pass on the current nightly: <https://rust.godbolt.org/z/Taxev5eMn>

library/core/src/array/iter.rs

@@ -266,7 +266,7 @@ impl<T, const N: usize> Iterator for IntoIter<T, N> {
         Fold: FnMut(Acc, Self::Item) -> Acc,
     {
         let data = &mut self.data;
-        self.alive.by_ref().fold(init, |acc, idx| {
+        iter::ByRefSized(&mut self.alive).fold(init, |acc, idx| {
             // SAFETY: idx is obtained by folding over the `alive` range, which implies the
             // value is currently considered alive but as the range is being consumed each value
             // we read here will only be read once and then considered dead.
@@ -323,6 +323,20 @@ impl<T, const N: usize> DoubleEndedIterator for IntoIter<T, N> {
         })
     }
 
+    #[inline]
+    fn rfold<Acc, Fold>(mut self, init: Acc, mut rfold: Fold) -> Acc
+    where
+        Fold: FnMut(Acc, Self::Item) -> Acc,
+    {
+        let data = &mut self.data;
+        iter::ByRefSized(&mut self.alive).rfold(init, |acc, idx| {
+            // SAFETY: idx is obtained by folding over the `alive` range, which implies the
+            // value is currently considered alive but as the range is being consumed each value
+            // we read here will only be read once and then considered dead.
+            rfold(acc, unsafe { data.get_unchecked(idx).assume_init_read() })
+        })
+    }
+
     fn advance_back_by(&mut self, n: usize) -> Result<(), usize> {
         let len = self.len();