feat(grafana): add performance dashboard

[lc525]

New dashboard showing:

• Filter-able to a given set of models/inference server pods:
  • per (model, inference server pod) throughput and average latency
  • aggregated per model throughput
  • aggregated per inference server pod throughput
• Filter-able to a given set of inference server pods
  • latency heatmaps (configurable rate interval) . agent -> inference srv -> agent . inference srv -> model -> inference srv
  • in-flight inference requests
  • CPU usage

Which issue(s) this PR fixes: Fixes #INFRA-1141 (internal) - performance dashboard