Figure out a way of having the per-pod latency graphs show per-cluster when multiple clusters are shown

[tomwilkie]

See #188 (review)