review: verdigris patina, honest Lane figures, tests that can fail (#102)

Two-axis review found the shipped colour and three weak tests.

- --patina was a warm gold at the same hue family as --brass; the spec
  asked for a cool blue-green so an unhealthy Lane is unmistakably not
  ember. Now verdigris in both branches, with docs/design-system.md
  stating why the far side of the wheel is the point.
- A Lane pass that returns before computing its figures carries the
  previous pass's due count and gap forward instead of recording zeroes,
  and Checked rides beside Due so a stopped Lane is distinguishable from
  a quiet one.
- TestAdminPageWithoutAPollerSaysSo now separates the two causes it
  conflated, TestOwnerClearsReaderMarks seeds real counters so the
  clearing assertion can fail, and TestLaneStatus asserts Checked and
  the carry-forward.
- backend/AGENTS.md records the one deliberate owner comparison outside
  requireOwner and the carry-forward rule.

backend/AGENTS.md

@@ -224,3 +224,12 @@ Guidance for OpenCode (and Claude Code) working under `backend/`. See root `AGEN
   rather than zeroes. `main.newRouter` takes the reporter as an interface and
   converts a nil `*Poller` to a nil interface — a typed nil would make the page
   claim a poller exists.
+  The one owner comparison left outside `requireOwner` is in `index`
+  (`view.Owner = readerID == h.store.OwnerID()`): it gates a link, not an
+  endpoint, so it is a rendering decision a registration-time wrapper cannot
+  express — do not "unify" it into the gate.
+  A Lane pass that returns before computing its figures (refusal backoff,
+  sidecar down) carries the previous pass's due count and gap forward rather
+  than recording zeroes; a Lane that has never reached a pace renders no gap at
+  all. `Checked` next to `Due` is what separates a stopped Lane from a quiet
+  one, so neither figure may be dropped from the row.