Spike: Improve RedisHLL metrics aggregation performance

As an outcome of #295204 (closed) we understand we cannot remove the existing slow metrics. So instead we need to find the way to speed up aggregation logic.

Why is it working slow

When we aggregate metrics we need to find union of HLL between set of keys. When amount of keys becoming significant, Redis need more than 0.01s which is considered as a slow operation. More details how we aggregate keys in this comment

What can we do

We shall consider migration daily to be weekly