Add metrics for number of series queried

[yeya24]

What this PR does:

This PR introduces a new feature to keep track of number of series queried on Ingester per tenant and expose it as metrics. Number of series queried can be tracked via configurable time windows so that we can keep track of number of series queried for the past 2h, 12h or 24h. This can be useful to understand query pattern and resource usage.

Here are some key points of this implementation:

• HyperLogLog or HLL is used to keep track of the cardinality of queried series. This is a probabilistic data structure to estimate cardinality with minimum memory usage.
• Each tenant will maintain a circular buffer of HLLs. Each HLL keeps track of the number of series queried for a given time window. If the time window size is configured to be 30m, to calculate the usage for the past 2h, we just need to merge HLLs from the past 4 time windows and estimate the final cardinality.
• Users can configure different time windows to keep track of the metric. For example, 2h and 24h. With 30m time window, the circular buffer will have 48 HLLs in total. When calculating 2h metric, 4 HLLs are merged. When calculating 24h metric, 48 HLLs are merged. The actual implementation utilized caching as only the latest HLL is changing and previous HLLs' merged results are cached.
• The hash of queried series is used to be added to HLLs to estimate cardinality. Sampling can be enabled via ingester.active-queried-series-metrics-sample-rate to reduce resource usage as queries go to ingesters are likely to be repetitive
• Updating HLL with series hash is implemented using an async producer consumer pattern using channels. This avoids adding additional latency to the Ingester query path.

Which issue(s) this PR fixes:
Fixes #7151

Checklist

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

[alanprot]

Thanks for doing this! great work! LGTM