Binop (e.g., add) broadcasting for shape (N, N) and (N) resulting in nan

[ariaghora]

    mt_tensor *t1 = mt_tensor_alloc_fill(MT_ARR_INT(10, 10), 2, 1.0f);
    mt_tensor *t2 = mt_tensor_alloc_fill(MT_ARR_INT(10), 1, 1.0f);
    mt_tensor *t3 = mt_add(t1, t2);