Kernel: Use alignas in Heap::AllocationHeader

Instead of adding manual padding and only on x86-64 and AArch64, use
`alignas(16)` for all architectures to ensure that allocations are
aligned on a 16 byte boundary.

This was originally guarded behind a `ARCH(X86_64)` when x86-64 support
was originally added.

Kernel/Heap/Heap.h

@@ -26,11 +26,7 @@ class Heap {
 
     struct AllocationHeader {
         size_t allocation_size_in_chunks;
-#if ARCH(X86_64) || ARCH(AARCH64)
-        // FIXME: Get rid of this somehow
-        size_t alignment_dummy;
-#endif
-        u8 data[0];
+        alignas(16) u8 data[0];
     };
 
     static_assert(CHUNK_SIZE >= sizeof(AllocationHeader));