[APM] Consolidate various throughput calculations to a utility function

[ogupte]

We have many instances of const deltaAsMinutes = (end - start) / 1000 / 60; followed by value / deltaAsMinutes accross APM UI, these should all be using a shared function which defines this logic once.

These instances can be found in x-pack/plugins/apm/server/lib/*:

• observability_overview/get_transaction_coordinates.ts
• services/get_service_dependencies/index.ts
• services/get_service_instances/get_service_instance_transaction_stats.ts
• services/get_service_transaction_groups/get_transaction_groups_for_page.ts
• services/get_service_transaction_groups/merge_transaction_group_data.ts
• services/get_services/get_service_transaction_stats.ts
• services/get_throughput.ts
• transactions/get_throughput_charts/index.ts
• transactions/get_throughput_charts/transform.ts

Related: #89350

[elasticmachine]

Pinging @elastic/apm-ui (Team:apm)

[dgieselaar]

There's also the ES rate aggregation. Previously that was not a great option because it didn't support histograms but now that we can use _doc_count it might become a viable alternative again

[sorenlouv]

The rate agg sounds great! Would love to outsource this to ES. Less code for us and seems more solid.
This would only be for timeseries data and not for the overall averages, right?

[dgieselaar]

@sqren yes, i assumed this issue was also about rate per date histo bucket but not sure. Rate only is allowed as a child of a date histo agg, but maybe it can work outside of one as well, but the semantics are less clear. E.g. it would need a fixed start/end or it could take the min/max of a field as the range. Not sure if it's worth it though

[sorenlouv]

Rate only is allowed as a child of a date histo agg, but maybe it can work outside of one as well, but the semantics are less clear. E.g. it would need a fixed start/end or it could take the min/max of a field as the range. Not sure if it's worth it though

Yes, that's what I understood as well. Probably better to only use it for timeseries and keep calculating the averages manually like today.

On a related note I wonder if of this will be useful in tackling: #89497