Allow pre-exposure start events for retention metrics

Feature request

Problem

Retention metrics currently require the start event to occur after feature flag exposure. This means you can't use signup as the start event when the flag fires later in the user journey.

A customer wants to answer: "among users randomly assigned at a flag gate, did the test variant improve signup-anchored retention?" This is a valid causal comparison (both groups are randomized at the gate), but the engine returns 0/0 because of the post-exposure timestamp filter.

Context

No major competitor supports this but I couldn't find a reason why this could be an issue.

Ticket: https://posthoghelp.zendesk.com/agent/tickets/52643 (moved to PostHog: https://us.posthog.com/project/2/support/tickets/55470)

Debug info

- [ ] PostHog Cloud, Debug information: [please copy/paste from https://us.posthog.com/settings/project-details#variables]
- [ ] PostHog Hobby self-hosted with `docker compose`, version/commit: [please provide]
- [ ] PostHog self-hosted with Kubernetes (deprecated, see [`Sunsetting Kubernetes support`](https://posthog.com/blog/sunsetting-helm-support-posthog)), version/commit: [please provide]

The issue names no files, tests, or entry points. Start by locating the retention-metrics logic that applies the post-exposure timestamp filter, then determine how pre-exposure start events should be handled; done means signup-anchored retention returns a valid comparison for users randomized at a later flag gate.