Change min-step to 15s to show better detail.

[bboreham]

What this PR does:
$__rate_interval will be floored at 4x this quantity, so 15s lets us see faster transients than the previous value of 1m.

Which issue(s) this PR fixes:
Fixes #339

Checklist

• CHANGELOG.md updated

[pracucci]

Makes sense to me!

Unrelated from this PR: I've also noticed in the in the queries.libsonnet dashboard we've some queries with the rate time window hardcoded to [1m]. Should we change it to [$__rate_interval] like other dashboards?

[bboreham]

we've some queries with the rate time window hardcoded to [1m]

The main difference I'm aware of is that the window will get bigger as you zoom out with [$__rate_interval]. So spikes will get more rounded off. At small time ranges it will still hit the floor so no difference; at large time ranges [1m] is very spiky so a smoother picture is easier to read.

Empirically on my machine it jumps from 1m to 2m between 24h and 2 days, and both are pretty spiky, so I would agree we should change queries.libsonnet to [$__rate_interval] like other dashboards.

Incidentally I found Beorn wrote an explanation which suggests that interval was intended to tell Grafana the scrape interval, so this PR is bringing the world into alignment.