nested graph_sub_nodes and graph input/output

This MR replaces nested graph_sub_nodes with graph input and output nodes.

This is an example of nested graph_sub_nodes (the final merged graph has only 2 nodes)

from ewokscore import load_graph


def myfunc(name=None, value=0):
    print("name:", name, "value:", value)
    return value + 1


subsubgraph = {
    "nodes": [
        {
            "id": "subsubnode1",
            "task_type": "method",
            "task_identifier": "__main__.myfunc",
            "inputs": {"name": "subnode1", "value": 0},
        }
    ]
}

subgraph = {
    "nodes": [{"id": "subnode1", "task_type": "graph", "task_identifier": subsubgraph}]
}

nodes = [
    {
        "id": "node1",
        "task_type": "method",
        "task_identifier": "__main__.myfunc",
        "inputs": {"name": "node1", "value": 0},
    },
    {"id": "node2", "task_type": "graph", "task_identifier": subgraph},
]
links = [
    {
        "source": "node1",
        "target": "node2",
        "sub_graph_nodes": {
            "sub_target": "subnode1",
            "sub_graph_nodes": {"sub_target": "subsubnode1"},
        },
        "arguments": {"value": "return_value"},
    }
]
graph = load_graph({"nodes": nodes, "links": links})
graph.execute()

With graph input_nodes and graph output_nodes this becomes

from ewokscore import load_graph


def myfunc(name=None, value=0):
    print("name:", name, "value:", value)
    return value + 1


subsubgraph = {
    "graph": {"input_nodes": {"in": {"id":"subsubnode1"}}},
    "nodes": [
        {
            "id": "subsubnode1",
            "task_type": "method",
            "task_identifier": "__main__.myfunc",
            "inputs": {"name": "subnode1", "value": 0},
        }
    ],
}

subgraph = {
    "graph": {"input_nodes": {"in": {"id":"subnode1", "sub_node":"in"}}},
    "nodes": [{"id": "subnode1", "task_type": "graph", "task_identifier": subsubgraph}],
}

nodes = [
    {
        "id": "node1",
        "task_type": "method",
        "task_identifier": "__main__.myfunc",
        "inputs": {"name": "node1", "value": 0},
    },
    {"id": "node2", "task_type": "graph", "task_identifier": subgraph},
]
links = [
    {
        "source": "node1",
        "target": "node2",
        "sub_graph_nodes": {
            "sub_target": "in",
        },
        "arguments": {"value": "return_value"},
    }
]
ewoksgraph = load_graph({"nodes": nodes, "links": links})
tasks = ewoksgraph.execute()

This idea is that when you use subgraphs, you only need to know the input_nodes and output_nodes. You no longer need to know what happens in the subgraphs.