feat(backend)!: split Series from Bookmark, keeping the wire format flat

Series becomes a shared row keyed (site, series_id) owning title, cover,
canonical URL, kind, latest chapter and last-checked time (ADR-0003).
A bookmark keeps only progress, favourite, lifecycle bucket, updated_at.

Store.Upsert decomposes one flat body across both tables in one
transaction: client title/series_url/cover apply only when the series row
is new, then only the poll may change them (security boundary — the row
is shared and the values are scraped page content). Reads join series
back in, so GET/PUT emit and accept exactly the flat field set they did
before (ADR-0004), asserted by TestFlatWireFieldSet.

The poller walks Series instead of Bookmarks: one fetch per shared
series per due cycle, due queue ordered reader_count DESC then
latest_checked_at ASC, orphaned series never due and never deleted, row
still stamped before the fetch. Batch/stagger/interval unchanged.

Migration 0002 backfills series from existing bookmarks; verified by
TestMigration0002BackfillsExistingBookmarks.

backend/AGENTS.md

@@ -21,7 +21,15 @@ Guidance for OpenCode (and Claude Code) working under `backend/`. See root `AGEN
   container per test binary (`TestMain` -> `pgtest.Main`) and hands each test
   its own database (`pgtest.URL(t)`). A package whose tests touch the store
   must have that `TestMain`.
-- **Single-user store.** One `bookmarks` table keyed `<site>:<series_id>` (`asura`|`demonic`|`comix`|`kagane`|`novelfull`|`lightnovelworld`), with a `kind` column (`manga`|`novel`) splitting the two libraries. Sync **last-write-wins**. Schema and endpoint list in plan.
+- **Single-user store, two tables.** `series` keyed `(site, series_id)`
+  (`asura`|`demonic`|`comix`|`kagane`|`novelfull`|`lightnovelworld`) owns the
+  shared facts — title, cover, canonical URL, `kind` (`manga`|`novel`),
+  Latest Chapter, `latest_checked_at` — and `bookmarks` holds only what
+  differs between readers: progress, favourite, lifecycle bucket,
+  `updated_at`. Sync **last-write-wins**; the wire format stays flat
+  (ADR-0004). `Store.Upsert` decomposes one flat body across both tables and
+  enforces the ownership rule: client `title`/`series_url`/`cover` are
+  written only when the series row is new (ADR-0003).
 - **Endpoints:** `GET /bookmarks`, `PUT /bookmarks/{key}` (upsert; see `updated_at` rule below), `DELETE /bookmarks/{key}`, `GET /healthz` (no auth).
 - **Web UI:** same binary serve password-gated browser UI on second
   hostname — `GET /` (list, or login page when no session),
@@ -53,22 +61,25 @@ Guidance for OpenCode (and Claude Code) working under `backend/`. See root `AGEN
   network access, so `latest_chapter` stay fresh when user not
   browsing. Second, parallel signal — userscript keep own
   `maybeCaptureLatestOnSeriesPage`/`backgroundRefreshLatest` logic unchanged.
-  Two independent clocks: per-bookmark cooldown (`latest_checked_at` column,
+  Two independent clocks: per-series cooldown (`series.latest_checked_at`,
   enforced by `Store.DueForLatestCheck`'s WHERE clause) and wake interval.
-  Row stamped *before* fetch so broken series wait out full
-  cooldown instead of retrying every tick, and writes go through
-  `Store.Get` + `Store.Upsert` so new chapter never reorders list.
+  The poller walks **Series, not Bookmarks** — a series referenced by several
+  bookmarks is fetched once per cycle, and the due queue orders
+  `reader_count DESC, latest_checked_at ASC` (ADR-0003). Series row stamped
+  *before* fetch so broken series wait out full cooldown instead of retrying
+  every tick; found chapter written straight to the series row via
+  `Store.SetLatestChapter`, so a bookmark's `updated_at` — and the list
+  order — is never touched.
   Fetches use `bogdanfinn/tls-client` with Chrome profile as defence in depth
   against fingerprint-based blocking; any failure log and skip. kagane and
   novelfull sit behind Cloudflare JavaScript challenges the TLS client can't
   clear, so they are browser-only: fetched over CDP via `BROWSER_WS_URL`, and
   simply not polled when that's unset. See
   `docs/superpowers/specs/2026-07-26-server-latest-chapter-polling-design.md`.
-  Poller's `Store.Get` + `Store.Upsert` not wrapped in transaction, so
-  userscript `PUT` that commits between the two can get overwritten by
-  poller's stale re-read — reverting that read progress and, since stored
-  value now differs, moving `updated_at` and reordering list. Known,
-  accepted limitation for single-user deployment, not bug to fix.
+  The poller's series write is a single-column UPDATE
+  (`Store.SetLatestChapter`), not a read-modify-write of the whole bookmark:
+  it cannot revert read progress or move `updated_at`, so the old
+  stale-re-read race is gone with the Get+Upsert flow.
 - **`updated_at` drives list order, so moves only on real reading progress:** server apply its timestamp when row new or `last_chapter_num` changes, else keep stored value — favouriting series or recording newly published chapter must not reorder list. `PUT` therefore returns row **as stored**, clients must adopt that response rather than own payload. See `plans/2026-07-25-bookmark-list-favorites-design.md` §4.
 - **Lifecycle buckets:** `status` on each bookmark is `reading` | `archived` |
   `finished`, orthogonal to `favorite`. Archived and finished appear only in