[test] Randomized Simulation Testing

[sujayakar]

Since our TCP stack is deterministic, we can get really good coverage via randomized simulation testing.

We could have a pretty simple testing scenario (Alice sending some fixed data to Bob) and then psuedorandomly schedule both of the engines, inject faults like packet drops & reorders, etc. Then, we should be able to check that the data eventually makes it there.

This will be especially good coverage for when both sides of the connections are reading and writing concurrently, since there's non-trivial interactions between the sending and receiving halves (e.g. including ACKs for data received on writes going out).