Number of iterations should be treated per client, not "global"

[danielmitterdorfer]

Consider the following task:

    {
      "operation": "my-query",
      "clients": 10,
      "iterations": 100
    }

Current behavior: Each client will execute iterations / clients (= 100 / 10) = 10 iterations. This can be sometimes puzzling and is also inconsistent to how we treat time-period. When a user specifies time-period, each client will run for that time period (and not a fraction of it).

New behavior: Each client will execute 100 iterations.

As this behavior may break some benchmarks (by running more iterations than expected) we should check whether there is an easy way to warn users of this new behavior (the warning will be removed after some grace period).