Asynchronously retry update on congestion

We now get notifications when the output buffer empties, and we
already caught incoming RTT pongs, meaning we can now react at the
proper time to retry a congested update rather than use a timer.
2 files changed