Declarative cross implementation test suite tools

[pipermerriam]

What was wrong?

• lots of duplication between the trio and asyncio based test suites.
• lots of things that behave differently when they occur across different processes.

How can it be fixed?

I've been imagining this concept that I'll call an "Endpoint Driver". It would be a declarative, implementation agnostic mechanism for defining a sequential set of actions that should be taken on an endpoint, with one of those actions being something like just running python code.

This tool should allow us to write implementation agnostic tests allowing us to define our test suite once and run it across any number of different combinations like:

• both endpoints of the same implementation in the same process
• different implementations in the same process
• same/different implementations in different processes.

I suspect the tests will be much more readable too.

[pipermerriam]

server_driver = compose_driver(
    connect_until_connected_to(client_config),
    wait_for(SomeEvent).add_handler(lambda ev: assert isinstance(ev, SomeEvent)),
)
client_driver = compose_driver(
    connect_to_endpoint(server_config),
    broadcast(SomeEvent()),
)

Then there's probably a backend type component for how to run each endpoint type as well as something to manage doing things in their own processes.

[pipermerriam]

Done with #141 opening new issues to convert the rest of the test suite to use this...