Sanity benchmark tcp with latency and many peers

[evan-forbes]

per comment

we should be able to recreate the issue using tcp and no comet code, if not, then we know that we do see the issue in comet, therefore there is likely some odd interaction between comet and the tcp connection.

It would also be interesting to compare with quic

AC

• write a tcp benchmark that uses a similar setup to comet
• attempt to recreate the issue where we are unable able to utilize the expected amount of bandwidth when there is network latency and many peers.

[evan-forbes]

this is different from #1413 in that this issue is purely using tcp, not comet, but the data from both would be useful in the end analysis to determine if we actually could replicate the issue.