Admin dashboard: pages, Lane observability, poll pass log, per-Series intervention (#134) (#148)

Spec #134, all ten tickets. Closes #134.

## What ships

The admin surface becomes four bookmarkable addresses behind one nav row, and Lane observability stops dying with the process.

- **#138** `/admin` splits into Overview, Lanes, Readers, Series, each a real route with the active tab underlined.
- **#139** `poll_passes` and `poll_lanes` land as durable tables with their store surface.
- **#140** cross-Series admin read model, with the privacy boundary in the projection: the Reader id that raised a Latest Chapter never leaves the store package.
- **#141** the poller records exactly one pass row per exit, with a skip reason and outcome counts.
- **#142** Series list: eight hygiene filters, Site and Library narrowing, paging — all of it in the query string, so a filtered list is a bookmark.
- **#143** Overview: a three-state verdict line and a stats block where every non-zero figure links to the list that counts it.
- **#144** per-Series detail page, keyed by the `site:series_id` composite the rest of the system already uses.
- **#145** the Lanes page reads the database; the in-memory Lane state, `web.LaneReporter` and `latest.Status` are deleted.
- **#146** Forced Poll: *Check now* stamps `series.force_poll_at` and never commands the poller.
- **#147** pause and resume one Site's Lane, with a mandatory 1h/6h/24h expiry.

## Shape of the design

Two decisions carry the rest. **Commands go through the database, never at the poller**: both *Check now* and a Lane pause write a row the next pass reads, so they survive a restart and the whole surface stays testable with no poller running. And **pending is derived, never stored** — the request stamp being newer than the check stamp — which self-clears on the check stamp with no second write and no sweeper, because the check stamp is written before the fetch.

ADRs: `docs/adr/0012-persisted-lane-state.md`, `docs/adr/0013-commands-through-the-database.md`.

## Verification

`go test ./...` green on the merged base (`264839e`), all packages, Docker-backed. `gofmt -l` and `go vet` clean.

Every ticket was reviewed on both axes (`cr-spec` + `cr-standards`) before merge.

## Known, non-blocking

- **#143** the verdict ignores never-reported Lanes when other Lanes have reported, and the per-Site table lists Sites that have Series rather than the whole registry. The ticket prose asks for eight hygiene figures per Site; the design mock and the landed `.tbl.sites` grid both say six columns, and the mock won.
- **#146** two `SeriesPage` scans per press instead of a keyed read — `ponytail:`-commented in-tree with the upgrade path.
- **#147** a paused Site with no pass row yet renders no row and so no control, since the Lanes page lists Sites that have passed.
- **#141** a sibling browser Lane declining at the top of a pass records as `sidecar-down`. Specified deliberately; the later spec in this series settles it.

Reviewed-on: #148
Co-authored-by: Sulthan Zaki <sultankiki05@gmail.com>
Co-committed-by: Sulthan Zaki <sultankiki05@gmail.com>

docs/adr/0012-persisted-lane-state.md

@@ -0,0 +1,44 @@
+# ADR-0012: Persisted lane state
+
+Date: 2026-08-21
+Status: accepted
+
+Supersedes the in-memory lane snapshot carried by `latest`'s `LaneState`/`Status`
+and the `web.LaneReporter` seam (ADR-0010 wrote the durable rows this page now
+reads).
+
+## Decision
+
+The admin Lanes page stops reading the poller's in-memory Lane state and
+becomes a read of `poll_passes`/`poll_lanes` in Postgres. There is no
+`LaneReporter` interface: `web/admin_lanes.go` walks `store.LatestLanePasses()`
+into one row per Site and adds the window's outcome sums from
+`store.LanePassOutcomes()`. The `latest` package's `LaneState`/`Status` snapshot
+and its `web.LaneReporter` seam are deleted.
+
+The browser is a deployment configuration fact plus a reachability derived
+from the pass log: `BROWSER_WS_URL` set means "configured", and the browser is
+"reachable" unless a recent browser-Site pass inside `latest.RefuseBackoff` is
+a sidecar loss, a missing fetcher, or an interrupted read. A skip reason is
+the whole difference between a Lane resting and a Lane stuck: a skipped pass
+prints its sentence, and only an empty skip with Series due and none read
+draws the true-stall fault. Sleep skips never count toward `Attention`.
+
+## Why
+
+The old page lived on a poller snapshot. Because that state was in memory, a
+deploy erased it: the page read zeroes until a fresh pass ran, and browser
+reachability came through a reporter interface only a live poller could
+serve. Making the page answer from the database means a restart is complete
+the instant the store is up, the browser fact survives a poller restart, and
+a Lane that has not yet gathered figures shows a placeholder rather than a
+confident zero.
+
+## Constraints
+
+The poller still owns the writes: each pass exit records one row (ADR-0010),
+and a pass that returns before gathering figures carries the previous pass's
+numbers forward instead of recording zeroes. A skip is a stable wire string;
+`asleep` never counts toward `Attention`. When polling is switched off
+(`LATEST_CHAPTER_POLL_ENABLED` unset) the page must say so, and the browser
+statusline appears only when polling is switched on.

docs/adr/0013-commands-through-the-database.md

@@ -0,0 +1,60 @@
+# ADR-0013: Commands through the database
+
+Date: 2026-08-22
+Status: accepted
+
+## Decision
+
+Owner interventions are **facts about rows, never commands to the poller**.
+*Check now* (`POST /admin/series/{key}/poll`) writes one stamp —
+`series.force_poll_at`, unix ms, zero meaning never asked (the column landed
+in migration 0014) — and the poller's next pass reads it through
+`Store.DueForLatestCheck`. The control never signals the running process, so
+a request survives a restart, and the whole surface is testable with no
+poller running at all.
+
+**Pending is derived, never stored**: a Series is pending while
+`force_poll_at > latest_checked_at`. It self-clears with no second write and
+no sweeper because the check stamp is written *before* the fetch (the same
+"attempted" discipline as ADR-0010) — the first attempt ends the pending
+state whatever the attempt returns. There is no expiry: a request the Lane
+never reaches keeps ageing in the UI, and an old pending marker is itself the
+evidence that a Lane is stuck. Writing again re-stamps the request time; the
+write is idempotent.
+
+**Queue-jump rules.** A forced Series overrides exactly three gates in the
+due query: the rest cutoff, the Sighting-deferral clause and the finished-only
+bucket, and it sorts to the front of the queue
+(`ORDER BY forced DESC, reader_count DESC, latest_checked_at ASC`). It never
+overrides an empty `series_url` (nothing to fetch), the Bookmarks join (a
+Series no Reader holds has no consumer for the result), the Lane's refusal
+backoff, the sidecar-down skip, or the Lane's gap — the last three are
+poller-side gates the query cannot see and must not. The one pass-level gate
+a forced Series does open is the browser wake threshold: a human asking wakes
+a sleeping Chrome, where the thresholds exist to stop the machine waking
+itself for one unattended check. If the home machine is off, nothing happens
+and the request ages visibly, which is correct.
+
+Rejected: zeroing the check stamp as the force signal. It would corrupt the
+never-checked and stale counts the landing page exists to show, and make a
+pending marker impossible.
+
+## Why
+
+A stuck-looking Series previously waited for its turn in the Lane's hour, and
+there was no way to ask for one check sooner. A direct poller command would
+have been lost on every restart and untestable without a running poller; a
+row the poller already reads is neither. Deriving pending from the two stamps
+keeps the flag honest across restarts and makes the mechanism two column
+writes and three query clauses instead of a state machine.
+
+## Constraints
+
+- The finished-status clause the force flag overrides is today's Lifecycle
+  test; a later spec in this series deletes it wholesale rather than amending
+  it, so the clause stays as it stands.
+- The control is unconfirmed (it takes nothing away) and renders no
+  `.confirm-row`; it is hidden on a Series with no `series_url` and on an
+  orphan — the same pair the due query refuses to override.
+- The answer to a press is the freshly rendered row, so the figures describe
+  the state after the press.