AK: Put fence functions in the global namespace

In some places, we already didn't use an "AK::" prefix, most likely due
to ADL. Instead of having to rely on ADL, let's just make this explicit.

AK/Fences.h

@@ -21,3 +21,8 @@ static inline void atomic_thread_fence(MemoryOrder order) noexcept
 }
 
 }
+
+#if USING_AK_GLOBALLY
+using AK::atomic_signal_fence;
+using AK::atomic_thread_fence;
+#endif

Kernel/Arch/SafeMem.h

@@ -37,7 +37,7 @@ struct RegisterState;
             return expected; // exchanged
 
         // This is only so that we don't saturate the bus...
-        AK::atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
+        atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
     }
 }
 
@@ -55,7 +55,7 @@ struct RegisterState;
             return expected; // exchanged
 
         // This is only so that we don't saturate the bus...
-        AK::atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
+        atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
     }
 }
 
@@ -73,7 +73,7 @@ struct RegisterState;
             return expected; // exchanged
 
         // This is only so that we don't saturate the bus...
-        AK::atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
+        atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
     }
 }
 
@@ -91,7 +91,7 @@ struct RegisterState;
             return expected; // exchanged
 
         // This is only so that we don't saturate the bus...
-        AK::atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
+        atomic_thread_fence(AK::MemoryOrder::memory_order_acquire);
     }
 }
 

Kernel/Tasks/AtomicEdgeAction.h

@@ -69,7 +69,7 @@ public:
             }
         }
 
-        AK::atomic_thread_fence(AK::memory_order_release);
+        atomic_thread_fence(AK::memory_order_release);
 
         if (expected == 1 << 1) {
             last_ref_action();