Add singlethreaded fence intrinsics.

These new intrinsics are comparable to `atomic_signal_fence` in C++,
ensuring the compiler will not reorder memory accesses across the
barrier, nor will it emit any machine instructions for it.

Closes #24118, implementing RFC 888.

src/libcore/intrinsics.rs

@@ -139,6 +139,21 @@ extern "rust-intrinsic" {
     pub fn atomic_fence_rel();
     pub fn atomic_fence_acqrel();
 
+    /// A compiler-only memory barrier.
+    ///
+    /// Memory accesses will never be reordered across this barrier by the compiler,
+    /// but no instructions will be emitted for it. This is appropriate for operations
+    /// on the same thread that may be preempted, such as when interacting with signal
+    /// handlers.
+    #[cfg(not(stage0))]     // SNAP 5520801
+    pub fn atomic_singlethreadfence();
+    #[cfg(not(stage0))]     // SNAP 5520801
+    pub fn atomic_singlethreadfence_acq();
+    #[cfg(not(stage0))]     // SNAP 5520801
+    pub fn atomic_singlethreadfence_rel();
+    #[cfg(not(stage0))]     // SNAP 5520801
+    pub fn atomic_singlethreadfence_acqrel();
+
     /// Aborts the execution of the process.
     pub fn abort() -> !;