[FEATURE] Let a /code-review (or reviewer agent) session return a resumable ID for targeted re-review after fixes land

Problem

Today, getting an independent code review means starting a fresh session (deliberately — same session as the author isn't a real second opinion). That reviewing session produces findings, then its context is effectively gone: it isn't tied to anything the author's working session can reference later.

Once the author fixes the flagged issues, there's no way to ask that same reviewer to check its own findings against the new code. The options today are:

1. Manually keep the reviewing session's terminal/window open and paste the fix back into it, or
2. Start yet another fresh review, which re-derives everything from scratch and doesn't know which specific findings it's supposed to be checking.

Neither preserves the actual value of the original review: a reviewer that already did the analysis and knows exactly what it flagged and why.

Note: within a single session, Agent-spawned subagents already support this — a resume-by-ID mechanism can bring a previously spawned agent back with full context. The gap is that this doesn't extend across sessions/entry points: a review started as its own top-level session (e.g. a fresh CLI invocation, or a scheduled/cloud review) has no equivalent handle that a different session can use to resume it later.

Proposed behavior

• Starting a review (/code-review, or a dedicated reviewer agent) returns a stable session/agent ID.
• That ID can be handed to a different session — typically the original working session that requested the review — and used later to resume the reviewer with its full original context (the diff it reviewed, its reasoning, its exact findings).
• A "re-review" trigger against that ID asks the reviewer to check the current state of the code against each of its own findings and report per-finding outcomes (resolved / not resolved / partially), rather than re-running a full review from scratch.

This mirrors GitHub's "re-request review" — tied to a specific reviewer, not a generic re-scan — applied to agent sessions.

Why this matters

Preserves reviewer independence (the reviewer still doesn't fix anything) while closing the loop that currently requires manual session bookkeeping to verify fixes actually address what was found.