Atomic_Win32: Replace deprecated (and since been removed) barrier intrinsics

As of VS 15.7, these seem to have been removed. Given we shouldn't have
been using these for some time, just replace them with the standard
library equivalent.

This fixes building on Windows with VS 15.7

Source/Core/Common/Atomic_Win32.h

@@ -8,6 +8,7 @@
 
 #include <Windows.h>
 
+#include <atomic>
 #include "Common/CommonTypes.h"
 
 // Atomic operations are performed in a single step by the CPU. It is
@@ -64,8 +65,10 @@ inline T AtomicLoad(volatile T& src)
 template <typename T>
 inline T AtomicLoadAcquire(volatile T& src)
 {
-  T result = src;  // 32-bit reads are always atomic.
+  // 32-bit reads are always atomic.
-  _ReadBarrier();  // Compiler instruction only. x86 loads always have acquire semantics.
+  T result = src;
+  // Compiler instruction only. x86 loads always have acquire semantics.
+  std::atomic_thread_fence(std::memory_order_acquire);
   return result;
 }
 
@@ -78,7 +81,8 @@ inline void AtomicStore(volatile T& dest, U value)
 template <typename T, typename U>
 inline void AtomicStoreRelease(volatile T& dest, U value)
 {
-  _WriteBarrier();  // Compiler instruction only. x86 stores always have release semantics.
+  // Compiler instruction only. x86 stores always have release semantics.
+  std::atomic_thread_fence(std::memory_order_release);
   dest = (T)value;  // 32-bit writes are always atomic.
 }