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/rustllvm/RustWrapper.cpp

@@ -189,8 +189,10 @@ extern "C" LLVMValueRef LLVMBuildAtomicCmpXchg(LLVMBuilderRef B,
                                                failure_order
                                                ));
 }
-extern "C" LLVMValueRef LLVMBuildAtomicFence(LLVMBuilderRef B, AtomicOrdering order) {
-    return wrap(unwrap(B)->CreateFence(order));
+extern "C" LLVMValueRef LLVMBuildAtomicFence(LLVMBuilderRef B,
+                                             AtomicOrdering order,
+                                             SynchronizationScope scope) {
+    return wrap(unwrap(B)->CreateFence(order, scope));
 }
 
 extern "C" void LLVMSetDebug(int Enabled) {