Add solution for multiple retention times

We often have to split logs into several retention categories. The usual way is to use different indices/ILM policies/data streams for types of logs.

We should have a way of splitting logs by condition (like in Redis outputs in pipeline management). I don't know if there's a solution with Elasticseach-pipelines but we can definitely have it in our Elasticsearch outputs in Logstash. If both works, why not implement both?

Review how Elasticsearch outputs and Logstash pipeline management currently handle index selection and retention policies. Clarify whether the change should support conditional routing in Elasticsearch outputs, Elasticsearch pipelines, or both. Done should mean logs can be split into multiple retention categories by condition, with the supported scope and behavior documented.