If the `detailed_monitoring` variable is set to true, ec2 instance
detailed monitoring will be enabled.

Ansible defaults to False, despite documentation that doesn't state
this.  Boto then turns off EBS Optimization on instance types where this
should be impossible.
https://github.com/ansible/ansible-modules-core/pull/2326

Includes mounting ephemeral disks

elbs:
  - name: "{{ e_d_c }}-public"
    scheme: "internet-facing"
  - name: "{{ e_d_c }}-internal"
    scheme: "internal"

If an internal elb exists, we need private subnets

This is functionally identical to the service_subnets and assumes you've
defined a private_elb_subnet_{1,2} similar to the existing
public_subnet_{1,2} variables (and requires code added in edx-secure's edx_service)

The subnet if statement is horride, but registering that variable as a
temporary global wasn't working.

Services might not have elbs, might only have public ELBs or might have
internal ELBs with no subnets configured.  Try to catch these cases
before ansible/ec2 give a less useful error message.

It can now bring up instances as well as ASGs for a service.

Also update the VPC play to fix some bugs.
 - Add an EIP to the NAT.
 - Source/Dest checks need to be disabled for NAT instances.