max TPS test can't recover if a node goes out of sync

[marta-lokhova]

Looks like if any node goes out of sync, max TPS test won't recover on the next iteration, causing it to fail even if the network can sustain the TPS rate in question. This is due to the fact that we don't restart nodes from scratch on each iteration (IIRC this was a problem with k8s being too slow spinning up pods for large missions). We should handle this by either allowing the node to catchup in between iterations, or only restarting failed nodes with the latest state.