Feature Request: average time between two events

Feature request

Is your feature request related to a problem?

When running experiments, there's no native way to use "average time between two events" as a metric. Users who want to measure things like time-to-complete a task, time between activation steps, or time from first action to conversion have to resort to workarounds. This makes it hard to use time-based metrics as experiment success criteria without extra instrumentation.

Describe the solution you'd like

A new experiment metric type (or an extension of the existing Mean metric) that allows users to select two events – a start event and an end event – and automatically computes the average time elapsed between them per user, then compares that mean across variants statistically.

This would mirror the "average time to convert" already shown in funnel insights, but exposed as a proper experiment metric with statistical significance testing.

Describe alternatives you've considered

Custom property on end event: Calculate the time delta client-side and send it as a property (e.g. time_to_complete_ms), then use a Mean metric on that property. This works but requires extra instrumentation and isn't available for users who don't control the event schema.
Funnel metric: Funnels show average/median conversion time between steps, but this isn't surfaced as a comparable numeric metric in experiments.
tickets/58377

Debug info

No file or test is named. Start with ticket 58377 and the existing funnel-insights conversion-time behavior described here, then locate the experiment metric and Mean metric entry points. Done means the two-event elapsed-time metric is available for experiments and supports statistical comparison across variants.