When we have high demand for runners, we exceed the capacity of our host
fleet.
To handle this spiky demand, we route the requests to AWS VMs. We use
"alien" to identify the VM on the external cloud rather than on our own
hosts.
This PR uses the same logic as Ubi-on-AWS, which is used for PostgreSQL.
When you provide a private location and AWS vm SKU, the nexus assembler
provisions it on AWS.
Currently, it's controlled by a feature flag in the project. After we
test it, we will change the logic to route only the exceeded demand.
2217f1637b