feat(backup): add forgejo dump systemd timer + wire into runbook

Daily 'forgejo dump' (tar.zst) to /opt/backups/forgejo with 14-day retention,
via a oneshot service + timer. Maintenance section now documents both backup
layers (Codeberg push-mirror for git, dump timer for full state).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

README.md

@@ -267,9 +267,21 @@ No changes to this repo. In the new project:
 ## Maintenance / hardening
 
 - **Backups:** the `forgejo_data` volume is now the source of truth (repos, DB,
-  registry blobs). Run `docker compose exec forgejo forgejo dump` on a schedule,
-  and set a **push-mirror** on each repo to a cloud host (GitLab/GitHub) for
-  off-box git backup.
+  registry blobs). Two layers:
+  - *Off-box git:* a **push-mirror** of each repo to Codeberg (repo → Settings →
+    Mirror Settings → Add Push Mirror; HTTPS URL + a `write:repository` token).
+  - *Full dump:* the systemd timer in [`backup/`](backup/) runs `forgejo dump`
+    daily and keeps 14 days under `/opt/backups/forgejo`. Install on the VPS:
+    ```sh
+    sudo install -m 0755 backup/forgejo-backup.sh /usr/local/bin/forgejo-backup.sh
+    sudo cp backup/forgejo-backup.{service,timer} /etc/systemd/system/
+    sudo systemctl daemon-reload && sudo systemctl enable --now forgejo-backup.timer
+    sudo systemctl start forgejo-backup.service   # test once
+    ls -lh /opt/backups/forgejo                   # confirm an archive landed
+    systemctl list-timers forgejo-backup.timer    # confirm next run scheduled
+    ```
+    Copy the dumps off-box too (they're useless if the VPS dies). Add
+    `--skip-package` in the script to exclude registry blobs if dumps get large.
 - **Registry GC:** per-commit `:sha` tags accumulate. Enable Forgejo's package
   cleanup rules / run periodic GC so the VPS disk doesn't fill.
 - **Certs:** never delete the `caddy_data` volume — losing it re-issues every