[Proposal] Include compaction in exporterhelper

[pmm-sumo]

Is your feature request related to a problem? Please describe.

OpenTelemetry Collector Contrib's groupbyattrs processor can be also used for compaction (essentially putting records with matching Instrumentation Scope/Resource under single Scope/Resource)

Such capability is useful in at least two cases:

1. General optimization of data representation
   Even with using batching, records that have a matching Instrumentation Scope and Resource can be spread across separate ResourceSpans/ResourceLogs/ResourceMetrics. This might happen due to a number of reasons - e.g. having small input batches or due to fragmentation created by groupbytrace processor or loadbalancing exporter.
   By "compacting" data together using groupbyattrs processor, we avoid duplication of Attributes under several Resource objects and data is represented more efficiently

2. Reduction of export requests for Jaeger exporter
   Due to specific of Jaeger proto, the exporter creates a request for each ResourceSpans. If data comes fragmented, this might lead to a significant increase in number of requests which due to connection latency might further lead to dropping data. An example of that is discussed in opentelemetry-collector-contrib/#9009. For explanation on how groupbyattrs processor helps here, refer to this comment

Proposal

Perhaps the code used for grouping (and compaction) could be put into some opentelemetry-collector package and then compaction could be used in exporterhelper together with batching, as discussed in #4646

Alternative

If we conclude this problem is Jaeger specific, I would suggest to move the common code to some package in opentelemetry-collector-contrib and then use it in groupbyattrs processor and jaeger exporter

[pmm-sumo]

Wanted to bring this issue for consideration. It might be Jaeger specific but then I'm wondering if this could be useful for other exporters as well