An app's tabs are the studio's lists: everything told, every change, every run #95

/:owner/:name/issues, /pulls and /runs render as studio lists (everything told, every change, every run) whenever the repository is an app and the browser is in studio mode: one row per tell/change/run in its studio word, who has it, when it last moved, and the row itself is the one link (to the tell, the change or the transcript). No labels, milestones or branch names.

• Filtering is ?word=told|working|look|shipped|stopped, a link in the seg bar, never a form, so it's shareable; no word shows what's open and moving (told/working/look). waiting and planned fold into working for filtering purposes — one bucketing function, not a re-derivation of the word rules (those still live solely in tellWord/pullWord/runWord).
• The repository header now labels the Issues/Pulls tabs "Told"/"Look" in studio mode; dev mode and non-app repos are untouched.
• Each list is bounded to one page (25 rows) and computes words per row, the same accepted trade the pull list already makes for its own check summaries — paging itself is unchanged, just the existing list's cursor.
• ?mode=dev on any of the three still renders today's page exactly.
• Added docs/STUDIO.md section "The three lists" and test/studio.test.ts coverage for all three lists, their word filters, ?mode=dev, and a non-app repository being untouched.

Not done: no per-word counts on the filter pills (would need extra grouped queries I judged not worth it for this pass), and the "shipped"/"stopped" word buckets are computed by fetching the closed-state page rather than a dedicated indexed query — same honest tradeoff the existing pull list already lives with.

bun x tsc --noEmit and the full check suite pass; the 5 failures seen in apps.test.ts/pulls.test.ts/cross-repo-pull.test.ts/web.test.ts are pre-existing sandbox issues (git auth/spawn), confirmed present on the base branch before my changes too.

Closes #56