add autoscaler for the ruler

[kd7lxl]

What this PR does:
Defines an optional HorizontalPodAutoscaler for the ruler.

Requires a config like:

config:
  ruler:
    enable_sharding: true
    ring:
      kvstore:
        store: memberlist
ruler:
  autoscaling:
    enabled: true

as documented in https://cortexmetrics.io/docs/guides/ruler-sharding/.

Not sure if we should make any of these default or a condition on the HPA. Or maybe added a fail block to check for these conditions.

In my experience, both memory and CPU scale pretty linearly with ruler load, so I've included both as target metrics for the ruler HPA.

Which issue(s) this PR fixes:
Fixes #

Checklist

• CHANGELOG.md updated - the order of entries should be [CHANGE], [FEATURE], [ENHANCEMENT], [BUGFIX]