The EC2 plugin will auto-provision as many slaves as there are builds in the
queue divided by number of executors per slave. With builds of a sufficiently
long duration and sufficiently rapid firing of SCM triggers, this could result
in a large number of EC2 instances being launched.

For reasons of constraining the resource consumption on EC2, it is desirable to
implement a limit for auto-provisioning EC2 slaves. There should be at least a
per-job limit, but both a per cloud and per AMI limit make sense.