Feature request: batch-fetch pull request review comment threads for an explicit PR list

### Problem

Today, pulling **inline review threads** for **multiple PRs** means **N separate** `pull_request_read` calls (`get_review_comments` per PR). For agent workflows (team playbooks, review digests, distilling recurring conventions into docs or rules), that is slow, error-prone, and uses more API round-trips than necessary.

### Proposed solution (v1)

Add a tool (name up to maintainers) that accepts:

- `owner`, `repo`
- `pull_numbers`: array of integers (**bounded**, e.g. configurable max such as 20–50)
- Optional pagination parameters consistent with existing `get_review_comments` behavior

**Behavior:** For each PR number, return the **same structured data** as today’s per-PR `get_review_comments` (threads + metadata such as resolved/outdated), keyed by `pull_number`.

**Explicit non-goals (v1):**

- No semantic ranking, clustering, or “top comments” inside the server.
- No implicit full-repo scan; the caller supplies an explicit list (e.g. from `search_pull_requests` / `list_pull_requests` / git).

### Why this matters

- Documents a **single supported** pattern for agents: select PRs → **one batch read** → summarize offline.
- Reduces N+1 calls for engineering hygiene (guidelines derived from real review text).

### Acceptance criteria (suggestion)

- [ ] Documented **max PR count** and behavior for missing/404 PRs (partial success vs fail-all—pick one and document).
- [ ] Tests + `script/generate-docs` / toolsnap updates per repo conventions.

### Related issues

- Reply to review comments (still one thread at a time today): https://github.com/github/github-mcp-server/issues/1323
- Copilot suggestion changesets not in API (platform/API gap): https://github.com/github/github-mcp-server/issues/2235
- Companion (batch PR metadata for changelogs): https://github.com/github/github-mcp-server/issues/2302

### Note

We are happy to prototype a PR if maintainers agree on v1 scope (explicit `pull_numbers[]`, caps, and error semantics).