Depends on #146

Need a system to reproducibly spin up clusters across multiple regions for geo-distributed scale testing. Each region should support 3.3k workspaces across ~20 nodes (depending on limits). This infrastructure doesn't need to be fully automated -- may be facilitated by additional scripts outside of the existing scale utilities.

The regionally distributed clusters in (US-WEST, EU-WEST, AP-WEST) should live on private network with VPC interconnect.

Scripts' requirements:

• Ad-hoc: Spinning up ~20 nodes housing ~3.3k workspaces
• Long-lived: external provisioner
• Long-lived: workspace proxy

Precursor to distributed load generation, handled in this ticket.