Change the expected output of a regression test

but I think is a bit more meaningful now. We do not actually see the value of
`p->next` but at least it is consistent (before `p->next` was different from
`tmp.next`).

regression/memory-analyzer/pointer_to_struct_01/test.desc

@@ -2,8 +2,7 @@ CORE
 main.gb
 --breakpoint checkpoint --symbols p
 struct S tmp;
-tmp = \{ \.next=\(\(struct S \*\)(NULL|0)\) \};
-p = &tmp;
-p->next = &tmp;
+tmp = \{ \.next=\(\(struct S \*\)0\) \};
+p = \&tmp;
 ^EXIT=0$
 ^SIGNAL=0$