Improve instance resource management

[rpappalax]

ardere assumes you need all the instances for each step up front, which isn't true if you're running sequential steps back-to-back and simply want to run a different set of tests using the same instance cluster repeatedly.

For example, if I have 3 steps that require 1, 2 and 3 attack nodes (per step) respectively, the broker sets up 6 nodes up front. This isn't a big deal for small attack swarms, but can get close to AWS instance quotas for larger swarms if we don't manage this more efficiently.

see releated:
Issue #47
Issue #75