fix(tui): TaskCreate/TaskUpdate render nothing inline, like TodoWrite

[agentforce314]

What

The task checklist HUD (#827) shipped, but the TaskV2 mutations kept printing their
raw results underneath it. Planning a turn therefore cost a screenful of JSON that
restated the checklist — one ⏺ TaskCreate(…) ⎿ {"task":{"id":…}} row per task up
front, then one ⏺ TaskUpdate ⎿ {"success":true,…,"statusChange":{…}} row per status
flip. TodoWrite was already suppressed for exactly this reason (the pinned checklist
is these tools' entire UI); TaskV2 is the same tool in a per-row shape, so it now gets
the same treatment. recordTodos() has already fed the HUD from the very call whose
row is dropped.

Suppression is by outcome, not just by name. tasks_v2.py answers a refused
mutation with a successful tool call carrying
{"success": false, …, "error": "Task not found"} (_task_update_call), and the HUD
then redraws the unchanged checklist — hiding that row would leave the miss with no
trace anywhere in the UI. So isChecklistHudOnly() consults the result text, and a
refused (or errored) mutation keeps its row.

TaskList / TaskGet / TaskOutput keep their rows unconditionally: they are reads
whose output (task bodies, background-shell logs) the HUD never shows.

The live spinner row in ToolTrail gets the same filter, so a mutation that will
render no completion line doesn't blink a row on the way either.

Changes

• ui-tui/src/lib/todo.ts — isChecklistHudTool() (TodoWrite/TaskCreate/TaskUpdate),
  checklistMutationRefused() (JSON success:false, with a prose fallback for
  pre-JSON backends) and isChecklistHudOnly().
• ui-tui/src/app/turnController.ts — recordToolComplete swaps the hardcoded
  name !== 'TodoWrite' check for isChecklistHudOnly(name, error, resultText).
• ui-tui/src/components/thinking.tsx — ToolTrail filters the live rows by the
  same predicate.

Tests

5 new vitest cases (todoSurface.test.ts, lib/todo.test.ts): suppression of
TaskCreate+TaskUpdate with the HUD still fed and no stranded spinner, a refused
TaskUpdate keeping its row, TaskList keeping its row, plus units for both predicates.
Suite: 1750 passed, 8 failed — the 8 failures are pre-existing on main@5441d5d
(verified by stashing this diff and re-running them: identical failures).

Verification — deepseek-v4-pro, tmux-driven TUI

Same prompt, same model, same workspace; only the branch differs.

BEFORE (main@5441d5d)

 ❯ Use only the task tools. First call TaskCreate four times to add these tasks: Map the parser, Port the lexer, Wire the CLI, Write tests. Then
   call TaskUpdate to mark the first one completed and the second one in_progress. Do not use any other tool, do not read or edit files. Reply DONE
 ⏺ TaskCreate(Map the parser)
   ⎿  {"task": {"id": "70efc8b96b67", "subject": "Map the parser"}}
 ⏺ TaskCreate(Port the lexer)
   ⎿  {"task": {"id": "c3d1abf634ca", "subject": "Port the lexer"}}
 ⏺ TaskCreate(Wire the CLI)
   ⎿  {"task": {"id": "3e90801481f7", "subject": "Wire the CLI"}}
 ⏺ TaskCreate(Write tests)
   ⎿  {"task": {"id": "8e395e86b773", "subject": "Write tests"}}
 ⏺ TaskUpdate
   ⎿  {"success": true, "taskId": "70efc8b96b67", "updatedFields": ["status"], "statusChange": {"from": "pending", "to": "completed"}}
 ⏺ TaskUpdate
   ⎿  {"success": true, "taskId": "c3d1abf634ca", "updatedFields": ["status"], "statusChange": {"from": "pending", "to": "in_progress"}}

 ⏺  DONE

 ▾ 4 tasks (1 done, 1 in progress, 2 open)
   ✔ Map the parser
   ◼ Port the lexer
   ◻ Wire the CLI
   ◻ Write tests

AFTER (this branch)

 ❯ Use only the task tools. First call TaskCreate four times to add these tasks: Map the parser, Port the lexer, Wire the CLI, Write tests. Then
   call TaskUpdate to mark the first one completed and the second one in_progress. Do not use any other tool, do not read or edit files. Reply DONE

 ⏺  DONE

 ▾ 4 tasks (1 done, 1 in progress, 2 open)
   ✔ Map the parser
   ◼ Port the lexer
   ◻ Wire the CLI
   ◻ Write tests

AFTER — a refused mutation still renders (TaskCreate landed and is HUD-only;
the bogus TaskUpdate keeps its row)

 ❯ Call TaskCreate once with subject Map the parser. Then call TaskUpdate exactly once with taskId deadbeef0000 and status completed - that id is
   deliberately wrong and I want to see what the tool returns. Do not call any other tool. Reply DONE.

 ⏺ TaskUpdate
   ⎿  {"success": false, "taskId": "deadbeef0000", "updatedFields": [], "error": "Task not found"}

 ⏺  DONE

 ▾ 1 task (0 done, 1 open)
   ◻ Map the parser

[github-actions]

Test Results

     4 files     966 suites   26m 58s ⏱️
14 634 tests 14 617 ✅ 17 💤 0 ❌
29 946 runs  29 891 ✅ 55 💤 0 ❌

Results for commit 7ebbd99.