Extend Jupyterhub deployment via the Kubernetes ServiceMesh implementation

The aim is to run user workloads across different physical sites from one platform. This could be with Jupyterhub user profiles to start with to run user pods on different sites, before we build in the intelligence to pick a location based on workflow requirements.

Anticipate Magenta team lead, with support from Coral (SPSRC?). Involved sites being in similar time zones will make the collaboration easier, large physical distances between the involved sites is not of particular importance at this stage since we are looking at functionality and feasibility, not performance. 

This will entail 

• physically distanced, independently deployed k8s planes accessible to each other (may be achieved by being externally accessible)
• install linkerd with shared trust anchors on both clusters
  • check proxy injection
• install multi cluster extension on both
• enable exporting the hub service to create a service mirror, add traffic split object
  • this will show load balancing
• Add few different user profiles that will map to pods being spun up at different physical sites
  • Could be achieved by exporting the user-scheduler service or adding node taints/tolerations (likely need to do both)
  • this is also a thing but a jupyterhub specific thing and not sure how it would work for non-commercial clusters https://multicluster-kubespawner.readthedocs.io/en/latest/index.html