[Story]: Retained-code and churn counterweight

## Summary
Build the differentiator none of the surveyed competitors (DX, Jellyfish, LinearB, Faros) ship: git-history duplication and refactor-ratio tracking correlated with AI-usage windows. Rising acceptance rate alongside rising code duplication is a quality regression, not a win, and no vendor-reported acceptance metric alone can distinguish the two.

## Intent / Source of truth
GitClear's 2026 finding, cited as the grounding data point: copy/paste climbed from 9.4% to 15.7% of changed lines, block duplication rose 81% since 2023, and moved/refactored code fell from 21% to 3.8%. 2026-08-13 backlog consolidation, competitive positioning review.

## Scope
- [ ] Static analysis over git history to compute duplication and refactor-ratio metrics per repo, per time window
- [ ] Correlation view joining these metrics against AI-adoption/acceptance windows (from stories 2-5's data)
- [ ] Flag windows where acceptance rises alongside duplication as a distinct signal, not folded silently into a generic "AI usage went up" narrative

## Out of scope
- Any vendor API work — this needs repo access and static analysis over git history, no vendor API involved
- Automated remediation/refactoring suggestions — this story is observability only

## Verification
A test repo with a known period of rising duplication shows the correlation view flagging it distinctly from periods of healthy refactor activity.

## Risk assessment
Needs read access to customer git history, which is a more sensitive data-access footprint than the API-based connectors — scope repo access carefully and document what is retained vs. computed transiently.

## 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.