Auto merge of #98755 - nnethercote:faster-vec-insert, r=cuviper

Optimize `Vec::insert` for the case where `index == len`.

By skipping the call to `copy` with a zero length. This makes it closer
to `push`.

I did this recently for `SmallVec`
(servo/rust-smallvec#282) and it was a big perf win in
one case. Although I don't have a specific use case in mind, it seems
worth doing it for `Vec` as well.

Things to note:
- In the `index < len` case, the number of conditions checked is
  unchanged.
- In the `index == len` case, the number of conditions checked increases
  by one, but the more expensive zero-length copy is avoided.
- In the `index > len` case the code now reserves space for the extra
  element before panicking. This seems like an unimportant change.

r? `@cuviper`

Author: bors

library/alloc/src/vec/mod.rs

@@ -1379,9 +1379,6 @@ impl<T, A: Allocator> Vec<T, A> {
         }
 
         let len = self.len();
-        if index > len {
-            assert_failed(index, len);
-        }
 
         // space for the new element
         if len == self.buf.capacity() {
@@ -1393,9 +1390,15 @@ impl<T, A: Allocator> Vec<T, A> {
             // The spot to put the new value
             {
                 let p = self.as_mut_ptr().add(index);
-                // Shift everything over to make space. (Duplicating the
-                // `index`th element into two consecutive places.)
-                ptr::copy(p, p.offset(1), len - index);
+                if index < len {
+                    // Shift everything over to make space. (Duplicating the
+                    // `index`th element into two consecutive places.)
+                    ptr::copy(p, p.offset(1), len - index);
+                } else if index == len {
+                    // No elements need shifting.
+                } else {
+                    assert_failed(index, len);
+                }
                 // Write it in, overwriting the first copy of the `index`th
                 // element.
                 ptr::write(p, element);