Warp E2E Stress Testing

[aaronbuchwald]

Warp currently includes one very small e2e test to ensure that we can send and receive a single warp message on a network where two subnets talk to each other. We should expand on this to add a load test utilizing the load simulator.

The current load simulator pre-generates a list of transactions and adds them to a channel to satisfy the requirements for the TxSequence interface: https://github.com/ava-labs/subnet-evm/blob/v0.5.2/cmd/simulator/txs/tx_generator.go and issuing them through a tx "agent", which executes all of the transactions in its sequence until the channel has been closed: https://github.com/ava-labs/subnet-evm/blob/v0.5.2/cmd/simulator/txs/agent.go#L29.

We should use this to create independent workers that will send warp messages, aggregate signatures, and then deliver them.

Each worker should create TxSequenceA to send warp messages and TxSequenceB to deliver them to the other subnet.

1. Create a TxSeqeunceA (https://github.com/ava-labs/subnet-evm/blob/v0.5.2/cmd/simulator/txs/tx_generator.go) that issues a batch of N (could be 1) unique send warp messages on Subnet A (maybe use the blockchainID + address + nonce as the payload of the message)
2. Wait for them to be accepted and aggregate signatures for each warp message (note: we can supply a worker to the Agent that adds transactions to a queue as they get accepted to aggregate warp signatures and pass them off to TxSequenceB)
3. Once each warp signature has been generated, generate transactions and add them to TxSequenceB (which will have a separate agent waiting for transactions on the queue)

This should support creating multiple workers to generate a sequence of send/receive transactions on SubnetA and SubnetB. As the first load test, we should run this with only two different subnets, so that we can create multiple pairs of these agents and support running either one agent that sends messages from SubnetA to Subnet B or multiple back and forth, so that we can scale up the load.