Improve delta snapshot system by dynamically increasing and reducing stored sent snapshot cluster size to ensure we give ample time for an ack - these could get lost of we send a lot of clusters per second. Another improvement is to predict last receive snapshot data and assume they will go through, optimizing by not sending the same value 50 times a second - if not ACK'd in time, we'll clear the prediction and send again. I'm seeing good results on TTT where the host goes from sending 1mb/s to ~18-30kb/s with one connected client in some cases. Add SteamEnums.RecvMaxMessageSize and SteamEnums.RecvBufferMessages. Set these values in bootstrap to higher values.