Support per-iteration sets of agents

[RobertKrawitz]

This is an RFE to support a different set of agents per iteration. At present, if I want a different set of agents, I need a different run. There is a possible separate RFE to support dynamic agents within an iteration, but that's beyond the scope of this although I'll mention a possible use case.

Example use case: running pod vertical scaling tests on Kubernetes using Kata Containers (each pod in a separate VM; the VMs are not persistent, their lifetime is that of their guest pod). This kind of test increases the number of pods per run until failure occurs (or we reach the desired number of pods). With pods running inside a VM, I want to deploy an agent inside each VM to collect data within the guest, in addition to one agent for the hosting node.

At present, I have to perform a separate run for each number of pods I'm testing, which will become unwieldly. Being able to perform each number of pods (I typically use a sequence of 1, 4, 10, 20, 30...) as a separate iteration within one run would make for a much tidier result set. In this case, the lifetime of a run would be greater than the lifetime of any of the VMs inside the run.

A possible use case for dynamic agents within an iteration would be if I wanted to perform this kind of test by simply increasing the number of pods, rather than deleting and re-creating them.