[observability] Show workspace success rate by cluster & include prebuilds

[kylos101]

Is your feature request related to a problem? Please describe

While working this issue, we found that when prebuilds fail, it reduces the workspace success rate. Why? Prebuilds are included with the grpc_server_handled_total metric, which is a component of our calculation:

sum(rate(grpc_server_handled_total{grpc_method="StartWorkspace",grpc_code!~"OK|ResourceExhausted"}[1d])) OR on() vector(0)
    /
    sum(rate(grpc_server_handled_total{grpc_method="StartWorkspace"}[1d]))

Additionally, we found it was difficult to view success rate by cluster (now that we have 4).

Describe the behaviour you'd like

1. Add a new pane that has one success rate line for each cluster, alternating the legend color for clusters
2. Remove the prebuild filters from the stop measures, because the start metric includes prebuilds too (many failed prebuilds this week caused the success rate to plummet).

i.e.

    sum(rate(gitpod_ws_manager_workspace_stops_total{reason="failed"}[1d])) OR on() vector(0)
    /
    sum(rate(gitpod_ws_manager_workspace_stops_total[1d]))

3. Avoid treating prebuilds which users cancel manually as failures. This will require some research, as its unclear if the status to exclude is Canceled or Aborted or something else. This is a guess, you'll need to read code and/or test to confirm. The existing promQL excludes OK and ResourceExhausted.

Describe alternatives you've considered

Update the grpc_server_handled_total metric to include a field that allows us to filter by workspace type, and exclude prebuilds. I opted to not go down this path, because we should pay more attention to prebuilds.

Additional context

This will give us a better picture of the current state for all workspaces, not just regular workspaces.