add tests for hole punching

[marten-seemann]

We should have tests to make sure that hole-punching through a NAT / firewall works. This test would mainly make sure that the DCUtR protocol is doing its job in coordinating the hole punch attempt.

This test would be especially valuable if it can be used to hole punch using different transports. That way, we can make sure that hole punching works on both TCP and QUIC, for example.

Main challenge: We need to find a way to model a NAT, so that we actually have something to punch through.

This test would build on (and probably replace) the Circuit v2 test (#116), similar to how the handshake RTT test is replacing the ping test.

[p-shahi]

Does this duplicate #21

If so let's close one or the other

[thomaseizinger]

Given that we don't even have #116 yet, does it make sense to start that? Now that rust-libp2p has full hole-punching support for QUIC as well, I'd really like some automated tests for that.

Main challenge: We need to find a way to model a NAT, so that we actually have something to punch through.

Similarly to how @mxinden used terraform for the perf tests, I think we could provision a network with a NAT on AWS and then deploy the nodes there and run the tests.

[marten-seemann]

Cross-posting here, I added some pointers for a dockerized NAT setup in #116 (comment).

[thomaseizinger]

For all interested parties, there is now a PR at #304.