Add optimized cluster overview dashboard

This optimized dashboard mainly lowers the cardinality of the CPU
metrics. Specifically instead of using `avg(rate(node_cpu_seconds_total`
which has a cardinality of total CPUs across all managed clusters, we
instead use `cluster:node_cpu:ratio` which has a cardinality of 1 per
cluster.

That is with 100 clusters, with 16 CPUs, the cardinality before was
100*16 = 1600, where as with this change we now only fetch 100 metrics.

This should scale quite a bit better on larger installations with many
clusters/nodes.

Signed-off-by: Jacob Baungard Hansen <[email protected]>