Mark remote read protocol and serving impl as stable.

Proposal

Our docs mentions remote read as not currently considered part of the stable API and is subject to change even between non-major version releases of Prometheus.. In practice this is not true, we never broke it, even when adding streaming mode. We are particularly careful when attempting any features.

I propose we mark the protocol and serving implementation stable. The motivation comes from DMs asking me if this protocol is really unstable and if they can use it in production. Obviously it's production ready, but some companies require this on paper. Perhaps do it on 3.x?

I am not sure about client implementation, I am personally not interested in maintaining/improving that use case--there are better tools to do aggregate from multiple sources than Prometheus (e.g. Thanos Querier, promxy), but maybe there are different opinions here. It feels we could achieve more deprecating this case for Prometheus (e.g. this blocked some PromQL engine optimizations in the past).

Acceptance Criteria

• We have consensus we want remote read protocol and our serving endpoint marked as stable.
• Stability updated in https://prometheus.io/docs/prometheus/latest/stability/ and https://prometheus.io/docs/prometheus/latest/querying/remote_read_api/

Start by reading the stability page and the remote read API documentation linked in the issue, then review the open discussion about the protocol, serving implementation, and client scope. Done means consensus on the stability change and updated wording on both documentation pages.