Type hint for return value of Stack's compute method should be NDArray rather than Tensor

[wong-1994]

TensorOperator's compute method is used to realize underlying cached_data, which is of Type NDArray. However, Stack's type hint imply that the return value is a Tensor.

class Value:
    """A value in the computational graph."""

    # trace of computational graph
    op: Optional[Op]
    inputs: List["Value"]
    # The following fields are cached fields for
    # dynamic computation
    cached_data: NDArray
    requires_grad: bool

class Stack(TensorOp):
    def __init__(self, axis: int):
        """
        Concatenates a sequence of arrays along a new dimension.
        Parameters:
        axis - dimension to concatenate along
        All arrays need to be of the same size.
        """
        self.axis = axis

    def compute(self, args: TensorTuple) -> Tensor:
        ### BEGIN YOUR SOLUTION
        raise NotImplementedError()
        ### END YOUR SOLUTION

    def gradient(self, out_grad, node):
        ### BEGIN YOUR SOLUTION
        raise NotImplementedError()
        ### END YOUR SOLUTION