Rewrite M_REDUCE to be much faster.

Rewrite M_SEQ to be much faster.
Change interface of M_SEQ as we can easily apply M_MAP / M_MAP2 to further apply macro.

tests/test-mcore.c

@@ -268,8 +268,8 @@ static void test_reduce(void)
 {
 #define id(x,a) a
 #define add(a,b) a + b
-  assert ( M_REDUCE2(id, add, d, M_SEQ(1, 5, M_ID)) == 5 * 6 /2);
-  assert ( M_REDUCE2(id, add, d, M_SEQ(1, 10, M_ID)) == 10 * 11 /2);
+  assert ( M_REDUCE2(id, add, d, M_SEQ(1, 5)) == 5 * 6 /2);
+  assert ( M_REDUCE2(id, add, d, M_SEQ(1, 10)) == 10 * 11 /2);
 }
 
 int main(void)