EC2 plugin fails to honour instance cap of 1

With a basic EC2 configuration of any AMI, if you set either or both the global instance cap and the AMI instance cap to 1, it will launch up to 2 nodes instead. If you omit a value from either, and rely on a cap only on the AMI or the global state, then the same occurs. 

This happens on 1.38 and 1.36. I haven't tested other versions.

This potentially affects instances that you rely on stopping instead of terminating, and also affects machines that rely on uniqueness, such as Active Directory joined clients.