GitLab

This was probably failing (rarely) due to the fact that we're shrinking the
timeout asynchronously (I think?).

mocknet: use peer ID in peer address

RandLocalTCPAddress is mostly useful when we *actually* want to listen on a real
address. Unfortunately, when running a bunch of tests, we can actually run
out.

With this change, a collision means we have a duplicate peer ID so yeah...

fixes #473

autorelay

also check that we don't initially advertise any.

macosx seems to fail intermittently, and that race is the likely culprit.