This PR lays the groundwork for index sampling in TSDB. To do so, it creates a new struct,

typeStatsstruct {Streamsuint64Chunksuint64Bytesuint64Entriesuint64}

A set of(user, from, through, matchers) will resolve to the approximation of underlyingstreams,chunks,bytes, andentries.

In the short term, this will enable dynamic sharding in the query planner. In the long term, we can expose this as an endpoint and use it to power a host of new features such as cardinality analysis and informing users how much data a query will touch before running it.

There will be a followup PR which wires this into the index-gateway grpc layer.

ref#5428