Support Cockroach DB multi-region deployments in terraform and tanka configs

[callumdmay]

As per the CRDB docs specifying a region in the locality flag is a prerequisite for creating a multi-region deployment. This is a step towards addressing #482

[linux-foundation-easycla]

• ❌ - login: @callumdmay / name: Callum May . The commit (9b25b7d, 262dc6d, cff6b09, f3dd016, 445420f) is not authorized under a signed CLA. Please click here to be authorized. For further assistance with EasyCLA, please submit a support request ticket.

[BenjaminPelletier]

/easycla

[BenjaminPelletier]

To check in on this one and summarize some offline conversations: my concern is that we want to make sure each DSS instance (the unit of deployment hosted by an organization) is prioritized to receive a complete replica of the DAR (i.e., Company 1's DSS instance should receive a full replica, Company 2's DSS instance should receive a full replica, Company 3's DSS instance should receive a full replica). Since CRDB attempts to spread replicas most widely at the highest level of locality, putting geographical region (region=${var.crdb_region}) before DSS instance identity (zone=${var.crdb_locality}) means that this change will make it less likely for each DSS instance to receive a full DAR replica (replicas will be spread across regions rather than DSS instances). If the important thing is to use the region key in the locality, it seems like we could simply replace zone= with region=, but I'm not sure that's important because the simplicity of our deployment structure means we can simply use explicit low-level replication constraints rather than the higher-level feature of multi-region capabilities. If we do want both levels of locality hierarchy (DSS instance + geographical region), the DSS instance should be the top level of hierarchy rather than geographical region.