metrics aggregation

[rade]

We need a way to aggregate metrics, so we can show metrics in higher-level views, e.g. pods, services.

See also #1257.

[rade]

One challenge with metrics aggregation - or even just single metrics on their own - is that being told "X is using 50% CPU" is rather meaningless since it gives no indication whether that is because the remaining 50% are used by some other tasks, or available. @paulbellamy and I were therefore wondering whether we should instead report on available resources. Headroom might be an even better option.

[rade]

@rndstr your work in #2664 et al will resolve this.