Implement bounded review execution for oversized or context-heavy changes

## Problem
`shared/policies/large-pr-partitioning.md` (deterministic diff-size threshold, partition construction, per-partition review, cross-partition aggregation/dedup) and `shared/policies/review-stopping-criteria.md` (the `REVIEW INCOMPLETE` contract) together satisfy nearly all of this issue's acceptance criteria. Both Skills' runbooks are wired to activate partitioning, and both are covered by regression tests (`tests/unit/review/test_large_pr_partitioning.py`, `tests/unit/review/test_review_stopping_criteria.py`).

The one remaining gap (re-verified by reading both files in full): `skills/github-pr-review/policies/reviewer-delta-review.md` and `policies/pr-scope.md` never reference `large-pr-partitioning.md`. `reviewer-delta-review.md`'s "Escalating from delta to full review" section lists several escalation triggers but none based on delta size, and there is no test or doc statement establishing whether a reviewed-SHA delta re-review that itself exceeds the partitioning threshold gets partitioned too, or is assumed always-small.

## Goal
Define and document how reviewed-SHA delta re-review interacts with the partitioning threshold, and add regression coverage for the chosen behavior.

## Scope
- Decide: does an oversized delta get partitioned using the existing `large-pr-partitioning.md` mechanism, or does exceeding the threshold force escalation to a full (and then possibly partitioned) review?
- Document the decision in `reviewer-delta-review.md` (cross-referencing `large-pr-partitioning.md`) and `pr-scope.md` as needed.
- Add a regression test proving the chosen behavior for a delta that exceeds the partitioning threshold.

## Non-goals
- Rebuilding or redesigning the large-change partitioning architecture, which already exists and works.
- Any change to normal full-review or normal-size delta-review behavior.

## Acceptance criteria
- [ ] `reviewer-delta-review.md` explicitly states how an oversized delta is handled, cross-referencing `large-pr-partitioning.md`.
- [ ] A regression test exercises a delta re-review that exceeds the partitioning threshold and asserts the documented behavior.

## Dependencies
Relates to #88, #89 (partitioning strategy already defined there). Parent: #48.

## Validation
- New/extended unit test alongside `tests/unit/review/test_large_pr_partitioning.py`.

Start with shared/policies/large-pr-partitioning.md and reviewer-delta-review.md, then inspect tests/unit/review/test_large_pr_partitioning.py and the related regression tests. Determine the behavior for an oversized reviewed-SHA delta before updating reviewer-delta-review.md and policies/pr-scope.md as needed. Done means the decision is cross-referenced and a regression test proves the behavior.

Written by the indexing model from the issue text.