feat(example): Add a python example to codegen whole gemm.

[KuangjuX]

resolved #16.

[haruhi55]

a typo here. "that utilizes all memory hierarchies"

[KuangjuX]

fixed.

[haruhi55]

I just have some questions due to my curiosity.

1. Does Tensor(data) also represent a node in the graph, functioning similarly to a "buffer"?

2. Intuitively, gemm is an operation with input and output. How does it differentiate between the input and output?

[KuangjuX]

1. A Tensor is essentially a Buffer, which I have renamed:

#[pyclass(unsendable, module = "buffer", name = "Tensor")]
pub struct PyBuffer(pub Rc<Buffer>);

2. gemm takes three parameters: a and b as inputs, and c as the output:

    #[staticmethod]
    fn gemm(a: PyRef<PyNode>, b: PyRef<PyNode>, c: PyRef<PyNode>) -> Self {
        let access_map = AccessMap::new(0, vec![]);

        let node_a = Rc::clone(&a.0);
        let node_b = Rc::clone(&b.0);
        let node_c = Rc::clone(&c.0);

        let gemm = Gemm::new(vec![node_a, node_b], node_c, Rc::new(access_map));

        let node = ThrillerNode::new(ThrillerNodeInner::Op(Box::new(gemm)));

        PyNode(Rc::new(RefCell::new(node)))
    }

[haruhi55]

I see. Thank you.

[haruhi55]

EdgeA_gemm and EdgeB_GEMM have different naming style, is this intentional?

[KuangjuX]

fixed.

[haruhi55]

LGTM