benchmarking: Run benchmarks with non default shard settings

[marclop]

Description

Currently, our benchmarking doesn't tweak any of the default integration settings or the Elasticsearch settings that would allow more writes to be performed concurrently and increase the event throughput.

I would be interesting to understand what the effect of tuning the index.number_of_shards has and what the its scaling properties are, since oftentimes, our customers want more processing power and they just increase the APM Server size, which provides a greater buffer, but in practice, doesn't increase the indexing throughput of Elaticsearch if that is the limiting factor with the default index.number_of_shards: 1.

Related issues

Follow up from #7858