feat(dashboard): show tickets as a full page instead of a rail tab (#1144)

[brillout]

Summary

• Tickets is now a full-width page instead of a tab in the 27rem right rail, reached from a "Tickets" row in the sidebar below Overview — no project needs to be selected first.
• The page is cross-project: every registered project's tickets/*.md backlog gets its own full-width section (new collectAllTickets/onAllTickets) — the table spans the whole pane rather than sharing it with grid columns — each with its own Import/Update-from-GitHub bar (the update button now sits adjacent to its "Updated from GitHub 2d ago" stamp, styled as an outline button).
• Each ticket row shows topics, spiked/planned, the effort its spike/plan recorded, status, then a trailing meta cluster of fixed-width, aligned columns in the order "Priority: n" (an empty slot when the ticket names none, so rows stay aligned), date ("2d ago"), and its GitHub: issue link (icon + #42, linking out — beside the row's button, so the row still opens the detail page).
• A "Sort by: Date/Priority" dropdown reorders the list (Date, the default, is the server's own newest-first order; Priority sorts highest-first, ties falling back to newest-first).
• Ticket dates come from the ticket's own filename (YYYY-MM-DD_slug.md) rather than the file's mtime (which drifted every time an update-from-GitHub run touched the file), falling back to mtime only for tickets written before the dated-filename format.
• A ticket whose .spike.md (else .plan.md) records an effort estimate — the "Human intervention effort: low" line the spike format asks for — surfaces it as WorkspaceTicket.effort, shown as "Effort: n" on the row and in the detail meta.
• Open/Closed filter checkboxes (Open only, by default) filter the backlog by the Status: open/closed header (Add Status: open/closed to every ticket header #1230); a project with tickets hidden by the filter says so instead of offering an import for work that already exists.
• Clicking a row opens its own page at /{projectId}/tickets/{filename} (same slug as the file), showing the ticket's entire markdown — not just the head a list row reads. All of a ticket's meta sits below the description, in one order: date, priority (spelled out as "Priority: n", not a bare number — the format's priority is a 0-10 scale), the ticket's GitHub: link, then status, topics, spiked/planned, and effort. The Queue button lives here too, since the list row no longer has room for it.
• route.ts extended: /tickets is the cross-project list, /{projectId}/tickets/{slug} is one ticket's page, alongside the existing /settings.
• Merged repeatedly with concurrent upstream work on the same tickets surface (feat(dashboard): update tickets/ from GitHub without re-importing #1222 "update tickets from GitHub", ticket-format doc changes, and Add Status: open/closed to every ticket header #1230's new Status:/Priority:/GitHub: headers) and reconciled cleanly each time.

Closes #1144. Required for the demo video per the issue.

Test plan

• pnpm exec vitest run in packages/framework-dashboard — 555 tests pass.
• pnpm exec tsc --noEmit in packages/framework-dashboard — clean.
• pnpm test in packages/the-framework — 1405 tests pass, 1 pre-existing skip (readTicket/readTicketsMeta/collectAllTickets coverage, plus date/topics/status/github/effort parsing, filename-vs-mtime dating, and newest-first sort).
• pnpm exec vite build in packages/framework-dashboard — clean, onAllTickets registered in the telefunc shield.

---

Generated by Claude Code

[brillout]

CI failure investigated: the single failing test was binding a topic run re-homes it into the bound project... (#1122) in daemon-workspace.test.js, with ENOTEMPTY: directory not empty, rmdir '.../.the-framework'. That file is unrelated to this PR's diff (tickets list/detail page), and running it in isolation locally passed 5/5 — this reads as a filesystem-cleanup race under CI's parallel test load, not a regression from these changes. Requested a re-run of just the failed job.

---

Generated by Claude Code

[brillout]

Re-run failed the same way (ENOTEMPTY: directory not empty, rmdir '.../.the-framework' on the same test, daemon-workspace.test.ts:253). Worth noting: that test's own source comments already document CI-only flakiness for this exact scenario ("a real bind... on a loaded runner kills it before it records anything — the run then looks like it never spawned, which is the CI-only 1 !== 2 this test kept failing with"). This is a different symptom (an ENOTEMPTY during cleanup instead of that 1 !== 2), but the same root cause category: real subprocess + real filesystem timing against a loaded CI runner, in a test file this PR does not touch. Triggered a second re-run to see if it clears.

---

Generated by Claude Code

[brillout]

CI red on the ticket-format spec ships in the package (not materialized), with priority/topics (#684/#674) (src/tickets.test.ts) — but this fails on main itself too (main's own CI run 30201542698 at the same commit c7fb48f fails the identical assertion: TICKETING_FORMAT.includes('priority: low/medium/high/urgent')). A concurrent commit updated prompts/ticketing_format.md to the new Priority: 10-0 scale without updating this test's literal-string check. Unrelated to this PR's diff (I don't touch tickets.test.ts or ticketing_format.md) — will re-merge and re-run once the base recovers.

---

Generated by Claude Code

[brillout]

Base recovered — main's ticket-format test was fixed in #1228. Merged that in (now unrelated-CI-flake-free on my side too, confirmed by re-running the full suite locally: 538 dashboard tests + 1391 framework tests, all green) and pushed. Also worth noting: while checking, main's own CI hit the same daemon-workspace.test.js ENOTEMPTY flake I saw earlier on this PR — good confirmation it's a genuine environment/load-timing issue rather than anything diff-specific, and it looks like it's already tracked (tickets/2026-07-25_rehome-test-loses-bind-under-load.md).

---

Generated by Claude Code

[brillout]

@suleimansh I didn't read the code.