fix(review): align chat summary with review content

[ryzizub]

Description

The chat summary emitted by /review (and the review step in /build) does not align with the review output written to disk. Ordering differs, wording drifts, and no shared identifier links a summary bullet to the corresponding finding in the file(s). When a user wants to act on "item 3," nothing unambiguously maps it to the underlying finding.

This blocks follow-up workflows — selective fixes, interactive triage, posting to GitHub — that need to reference a specific finding.

Steps to Reproduce

1. Run /review on a branch with a non-trivial diff.
2. Compare the summary printed in chat against the review file(s) under docs/code-review/ (or docs/reviews/ during /build).
3. Observe: bullet order, titles, and counts differ; findings carry no IDs.

Expected Behavior

• Findings are generated once with a stable identifier (e.g., R-001, R-002).
• Chat summary and review file(s) render from the same source, sharing IDs, ordering, and titles.
• IDs are stable within a single review run so the user can say "apply R-003, skip R-007" unambiguously.

Acceptance:

• Every finding has a unique ID.
• Chat summary and review file(s) list findings in the same order, with matching IDs and titles.
• IDs are referenced in chat output so the user can act on them by number.

Additional Context

• Compatible with the companion issue on consolidating review output into a single file — IDs and alignment apply whether findings live in one file or several.
• Out of scope: severity tiers, interactive triage, GitHub comment bridge (separate issues).