Use QUIC's congestion window size when sending data

During synchronization, a peer starts sending fairly large buffers (64KB at the time of this writing) and that can create congestion.

It might be faster to start w/ a buffer size equal to the congestion window size of the peer we're syncing with.

It would be better to respond to backpressure in a more appropriate way. Essentially every time a chunk is not fully written, then we've surpassed the congestion window size (which can be modified).