feat: stabilize non-coordinated omission mode

[romange]

1. Our latency/RPS computations were off because we started measuring before drivers started running. Now, Run/Start phases are separated, so the start time is measured more precisely (after the start phase)
2. Introduced progress per driver connection - one of my discoveries is that driver connections progress with differrent pace when running in coordinated omission. This can reach x5 speed differrences. Now we measure and output the fastest/slowest progress.
3. Coordinated omission is great when the Server Under Test is able to sustain the required RPS. But if the actual RPS is lower than the one is sent than the final latencies will be infinitely high. We fix it by introducing self-adjusting sleep interval, so if the actual RPS is lower we will increase the interval to be closer to the actual RPS.

[romange]

factored out parser lib so that dfly_bench won't be dependent on facade

[romange]

this is an async function that only creates all the running fibers but does not run them

[chakaz]

It's not really async on its own, right? It just creates other async Fibers

[romange]

right

[romange]

moved from Run() function

[romange]

the only change besides the indentation is computation of min/max done metric per thread client.

[chakaz]

Many nits basically :)

[chakaz]

Nit

Suggested change

	if (done > max)
	max = done;
	if (done < min)
	min = done;
	max = std::max(done, max);
	min = std::min(done, min);

Or maybe even use:

    std::tie(min, max) = std::minmax_element(drivers_.begin(), drivers_.end(), [](const auto& a, const auto& b) { return a.done() < b.done(); }

But then min and max need to be iterations, and you'll return {min->done(), max->done()} below

[romange]

I chose the first version :)

[chakaz]

Maybe in the future you could count this (and below) as well

[chakaz]

It's not really async on its own, right? It just creates other async Fibers