make the dashboard multi-instance capable

[Dieterbe]

my vision (this is basically how i've been doing things since forever, but never really documented it):

1. dashboard.json in this repo is the canonical dashboard, it should have the latest and greatest stuff, and gets synced to grafana.com
2. our dashboards in our internal monitoring are as close as possible to the official dashboard.json
3. we try to keep up dashboard.json in this repo as we add new metrics to the code
4. periodically we integrate any changes we may have made internally (hopefully not too many) into source control, and then update the dashboard in our prod monitoring.

this time around, things were more interesting, as we have 2 internal dashboards, one per-instance, and one for multi-instance, which had some changes, but was started from an older 0.7.0-beta1 dashboard, so it also was missing a bunch of stuff.

i have done step 4 which was a bit more work now, but my idea is from now on we can use 1 dashboard that can serve both the single-instance, as well as multi-instance use cases.
it means making a few compromises here and there (in particular, making the legend useful both for many different instances as well as 1 specific instance, means it won't always look optimal for both), but I think it's worth it.

you can check out metrictank-0-7-4 in our internal monitoring to see it in action.
i already removed our older single-instance dashboard and plan to remove the metrictank-0-7-0-beta1-k8s-cluster one as well.

[Dieterbe]

@woodsaj @replay sound good?

[woodsaj]

This works for me. This ties in well with future hosted-metrics updates when users may have two instances show up in their metrics (eg, us-east and us-west)