Optimize vec::retain performance

This simply moves the loops into the inner function which leads to better results.


```
old:

test vec::bench_retain_100000                            ... bench:     203,828 ns/iter (+/- 2,101)
test vec::bench_retain_iter_100000                       ... bench:      63,324 ns/iter (+/- 12,305)
test vec::bench_retain_whole_100000                      ... bench:      42,989 ns/iter (+/- 291)


new:

test vec::bench_retain_100000                            ... bench:      42,180 ns/iter (+/- 451)
test vec::bench_retain_iter_100000                       ... bench:      65,167 ns/iter (+/- 11,971)
test vec::bench_retain_whole_100000                      ... bench:      33,736 ns/iter (+/- 12,404)
```

Author: the8472

library/alloc/src/vec/mod.rs

@@ -1520,49 +1520,46 @@ impl<T, A: Allocator> Vec<T, A> {
 
         let mut g = BackshiftOnDrop { v: self, processed_len: 0, deleted_cnt: 0, original_len };
 
-        // process_one return a bool indicates whether the processing element should be retained.
-        #[inline(always)]
-        fn process_one<F, T, A: Allocator, const DELETED: bool>(
+        fn process_loop<F, T, A: Allocator, const DELETED: bool>(
+            original_len: usize,
             f: &mut F,
             g: &mut BackshiftOnDrop<'_, T, A>,
-        ) -> bool
-        where
+        ) where
             F: FnMut(&mut T) -> bool,
         {
-            // SAFETY: Unchecked element must be valid.
-            let cur = unsafe { &mut *g.v.as_mut_ptr().add(g.processed_len) };
-            if !f(cur) {
-                // Advance early to avoid double drop if `drop_in_place` panicked.
-                g.processed_len += 1;
-                g.deleted_cnt += 1;
-                // SAFETY: We never touch this element again after dropped.
-                unsafe { ptr::drop_in_place(cur) };
-                // We already advanced the counter.
-                return false;
-            }
-            if DELETED {
-                // SAFETY: `deleted_cnt` > 0, so the hole slot must not overlap with current element.
-                // We use copy for move, and never touch this element again.
-                unsafe {
-                    let hole_slot = g.v.as_mut_ptr().add(g.processed_len - g.deleted_cnt);
-                    ptr::copy_nonoverlapping(cur, hole_slot, 1);
+            while g.processed_len != original_len {
+                // SAFETY: Unchecked element must be valid.
+                let cur = unsafe { &mut *g.v.as_mut_ptr().add(g.processed_len) };
+                if !f(cur) {
+                    // Advance early to avoid double drop if `drop_in_place` panicked.
+                    g.processed_len += 1;
+                    g.deleted_cnt += 1;
+                    // SAFETY: We never touch this element again after dropped.
+                    unsafe { ptr::drop_in_place(cur) };
+                    // We already advanced the counter.
+                    if DELETED {
+                        continue;
+                    } else {
+                        break;
+                    }
+                }
+                if DELETED {
+                    // SAFETY: `deleted_cnt` > 0, so the hole slot must not overlap with current element.
+                    // We use copy for move, and never touch this element again.
+                    unsafe {
+                        let hole_slot = g.v.as_mut_ptr().add(g.processed_len - g.deleted_cnt);
+                        ptr::copy_nonoverlapping(cur, hole_slot, 1);
+                    }
                 }
+                g.processed_len += 1;
             }
-            g.processed_len += 1;
-            return true;
         }
 
         // Stage 1: Nothing was deleted.
-        while g.processed_len != original_len {
-            if !process_one::<F, T, A, false>(&mut f, &mut g) {
-                break;
-            }
-        }
+        process_loop::<F, T, A, false>(original_len, &mut f, &mut g);
 
         // Stage 2: Some elements were deleted.
-        while g.processed_len != original_len {
-            process_one::<F, T, A, true>(&mut f, &mut g);
-        }
+        process_loop::<F, T, A, true>(original_len, &mut f, &mut g);
 
         // All item are processed. This can be optimized to `set_len` by LLVM.
         drop(g);