[Story]: Cross-vendor acceptance-rate normalizer

## Summary
Build one canonical cross-vendor acceptance-rate rollup with a documented per-vendor definition mapping. The core rule: never average raw vendor-reported percentages directly against each other — the denominators genuinely differ (ghost-text vs agent-mode, Amazon Q's continued-typing exclusion, Copilot's aggregate-only floor), so a naive average would be statistically meaningless.

## Intent / Source of truth
Epic finding + story 1's per-vendor denominator mapping. 2026-08-13 backlog consolidation.

## Scope
- [ ] Canonical rollup metric definition that normalizes on the schema from story 1 rather than on raw vendor percentages
- [ ] Documented per-vendor definition mapping surfaced alongside the rollup (so a viewer can see WHY numbers differ, not just that they do)
- [ ] Cross-vendor comparison view/dashboard

## Out of scope
- Building this before at least two connectors exist — there is nothing to normalize yet
- Hiding or collapsing the denominator differences in the name of a "clean" single number

## Verification
Dashboard showing at least two vendors' data side by side with the definition-mapping documentation visible; a spot-check confirming the rollup does not just average raw percentages.

## Risk assessment
DEPENDENCY: at least two connector stories (2-5) must land first — there is no normalizer without real data to normalize. The main product risk is producing a misleadingly "clean" number that hides real methodological differences.

## AI Usage Declaration
Drafted with AI assistance during the 2026-08-13 cross-repo backlog consolidation. A human owns intent, verification and consequences; acceptance criteria must be verified against the live system before this is closed.