[util] Force 32-bit windows malloc to be 16-byte aligned

By replacing it :P (and its friends). This gets the non-sw renderers
working with recent scene changes.

libs/scene/hierarchy.c

@@ -40,17 +40,6 @@
 
 #include "scn_internal.h"
 
-#if defined(_WIN32) && !defined(_WIN64)
-// FIXME (maybe) this is a hack to make DARRAY arrrays 16-byte aligned on
-// 32-bit systems (in particular for this case, windows) as the vectors and
-// matrices require 16-byte alignment but system malloc (etc) provide only
-// 8-byte alignment.
-// Really, a custom allocator (maybe using cmem) would be better.
-#define free(mem) _aligned_free(mem)
-#define malloc(size) _aligned_malloc(size, 16)
-#define realloc(mem, size) _aligned_realloc(mem, size, 16)
-#endif
-
 static void
 hierarchy_UpdateTransformIndices (hierarchy_t *hierarchy, uint32_t start,
 								  int offset)

libs/util/sys.c

@@ -646,6 +646,39 @@ Sys_PageIn (void *ptr, size_t size)
 //#endif
 }
 
+#if defined(_WIN32) && !defined(_WIN64)
+// this is a hack to make memory allocations 16-byte aligned on 32-bit
+// systems (in particular for this case, windows) as vectors and
+// matrices require 16-byte alignment but system malloc (etc) provide only
+// 8-byte alignment.
+void *__cdecl
+calloc (size_t nume, size_t sizee)
+{
+	size_t size = nume * sizee;
+	void *mem = _aligned_malloc (size, 16);
+	memset (mem, 0, size);
+	return mem;
+}
+
+void __cdecl
+free (void *mem)
+{
+	_aligned_free (mem);
+}
+
+void *__cdecl
+malloc (size_t size)
+{
+	return _aligned_malloc (size, 16);
+}
+
+void *__cdecl
+realloc (void *mem, size_t size)
+{
+	return _aligned_realloc (mem, size, 16);
+}
+#endif
+
 VISIBLE size_t
 Sys_PageSize (void)
 {