[WIP] lnprototest refactoring

[vincenzopalazzo]

We are embarking on an experimental journey with lnprototest.

Presently, lnprototest exhibits a certain level of intricacy in our test-writing approach.

Thus, this pull request (PR) introduces an idea that emerged from discussions on the core lightning Discord.

The essence of this idea involves transforming the runner into a stateless entity while endowing the connection with statefulness. By adopting this approach, we can embrace a more procedural manner of crafting tests (idea stoled from the PoC of Rusty https://github.com/rustyrussell/lnprotest).

Allow me to illustrate with an example:

def test_v2_init_is_first_msg(runner: Runner, namespaceoverride: Any) -> None:
    """Tests workflow
    runner --- connect ---> ln node
    runner <-- init ------ ln node
    """
    runner.start()

    conn1 = runner.connect(None, connprivkey="03")
    init_msg = conn1.recv_msg()
    assert (
        init_msg.messagetype.number == 16
    ), f"received not an init msg but: {init_msg.to_str()}"
    conn1.send_msg("init", globalfeatures="", features="")
    runner.stop()

Subsequently, we should be able to circumvent code duplication and compose code in a maintainable manner. To achieve this, we are adopting the framework concept proposed by Christian, available at: https://github.com/cdecker/lnpt.

@lnpt.step()
def connect(c):
    # Open a connection

@lnpt.step(parents=[connect])
def init(c):
    # Send and receive init

Furthermore, in conclusion, we can package all these developments into a library, eliminating the need to include one of the lnprototest repositories. This approach empowers each implementation to construct its own tests, a capability that already exists at present.

For in-depth discussions, please refer to: #105.

Note: I am intending to maintain backward compatibility with the old version of lnprototest initially, but after a year, I plan to discontinue support for it. The eventual removal of the old code is anticipated within two years or sooner.