Support benchmarking specific feature such as Tail Based Sampling

[marclop]

Description

We'd to not only benchmark the raw ingest performance of APM Server, but also build a framework that permits testing features that rely on the actual event trace.id to function correctly. With the upcoming benchmarking framework, we do not yet have a way to re-generate trace IDs on the fly, it simply replays the stored events as fast as possible.

The main feature we'd like to test is Tail Based Sampling, but there may be others in the future.

[simitt]

Let's focus on support for TBS in 8.3, marked it as a stretch goal.