Add dynamic load determination, e.g. based on latency, connection time, etc.

[peanball]

Hey can be used to stress test a server, but also to determine the correct sizing for a particular workload with a reasonably stable request pattern (requests & connections / s)

We already capture information on the various metrics, e.g. TLS handshake times, response latencies, etc.

By annealing to a desired target value for e.g. latency, the number of requests could be gradually increased until we reach the desired threshold.

During the gradual increase, the requests/s and other metrics can still be retained to check or improve the fine-tuning of a server.