More inference-friendly API for lazy

The signature for new was

```
fn new<F>(f: F) -> Lazy<T, F>
```

Notably, with `F` unconstrained, `T` can be literally anything, and just
`let _ = Lazy::new(|| 92)` would not typecheck.

This historiacally was a necessity -- `new` is a `const` function, it
couldn't have any bounds. Today though, we can move `new` under the `F:
FnOnce() -> T` bound, which gives the compiler enough data to infer the
type of T from closure.

library/std/src/sync/lazy_lock.rs

@@ -44,17 +44,14 @@ pub struct LazyLock<T, F = fn() -> T> {
     cell: OnceLock<T>,
     init: Cell<Option<F>>,
 }
-
-impl<T, F> LazyLock<T, F> {
+impl<T, F: FnOnce() -> T> LazyLock<T, F> {
     /// Creates a new lazy value with the given initializing
     /// function.
     #[unstable(feature = "once_cell", issue = "74465")]
     pub const fn new(f: F) -> LazyLock<T, F> {
         LazyLock { cell: OnceLock::new(), init: Cell::new(Some(f)) }
     }
-}
 
-impl<T, F: FnOnce() -> T> LazyLock<T, F> {
     /// Forces the evaluation of this lazy value and
     /// returns a reference to result. This is equivalent
     /// to the `Deref` impl, but is explicit.

library/std/src/sync/lazy_lock/tests.rs

@@ -136,6 +136,12 @@ fn sync_lazy_poisoning() {
     }
 }
 
+// Check that we can infer `T` from closure's type.
+#[test]
+fn lazy_type_inference() {
+    let _ = LazyCell::new(|| ());
+}
+
 #[test]
 fn is_sync_send() {
     fn assert_traits<T: Send + Sync>() {}