feat(spans): Collect exclusive time for all spans

[phacops]

We would like to start collecting the exclusive time for all spans while limiting cardinality by only tagging with low cardinality values. This is for @Zylphrex 's work with the metrics explorer.

[iker-barriocanal]

LGTM.

To validate exclusive_time_condition, could we add a test to ensure we don't emit more than one exclusive time metric for each span?

[jjbayer]

What about exclusive_time_light? Do we need the transaction tag for the metrics explorer? If yes, are there queries that would benefit from using exclusive_time_light (fewer rows to scan)?

[Zylphrex]

What about exclusive_time_light? Do we need the transaction tag for the metrics explorer? If yes, are there queries that would benefit from using exclusive_time_light (fewer rows to scan)?

@jjbayer So the metrics explorer wants to expose the ability to surface samples for metrics including this span exclusive time metric. One of the main things we want to be able to support is to query all spans hence the motivation to extract this metric for all spans instead of known module spans.

Now we want the transaction tag for the metrics explorer because it is likely one of the main ways for people to filter and narrow down the set of spans they want. To my understanding, the exclusive time light metric was specifically added as an optimization for the use cases where filtering on a transaction is not needed. There will be some queries that will benefit from using the exclusive time light metric so I will be raising this point to the team but we will still want this exclusive time metric for all spans.

[jjbayer]

There will be some queries that will benefit from using the exclusive time light metric

@Zylphrex alright, feel free to add it later in a follow-up PR!

[jjbayer]

See open question about duration_condition.

[jjbayer]

Should we still apply duration_condition here? Asking because even span_per_op and spans_per_segment apply the duration condition.

[phacops]

I don't think we'd want to apply the duration condition here. We want to track absolutely all spans here and the condition was mainly created for a mobile specific use case.