fix(cli): wrap AskUserQuestion options instead of truncating at terminal width (#4610)

[me2seeks]

Summary

When the agent asks a question via AskUserQuestion, option rows were hard-truncated at the terminal width — no wrap, no ellipsis, no way to read the tail. Users had to pick between protocol-compatibility strategies from half a sentence (see the issue screenshot).

UserQuestionOverlay now wraps option rows (and the question title) with pi-tui's ANSI-aware wrapTextWithAnsi: continuation lines indent under the option body, and the active row's highlight band covers every wrapped line. Because wrapping can outgrow the overlay's maxHeight — pi-tui clips overflow from the bottom, which would silently drop the free-text input row — the runner passes a live row budget (terminal.rows - margin, re-read per render so resize stays correct) and the overlay degrades gracefully when over budget: title capped at two lines, each option clamped to an equal share with a visible …, so the hint, every option, the input row, and the divider always render.

Fixes #4610

Verification

• packages/cli full suite: 757 tests, 0 fail — including 9 new tests in pi-tui-user-question-option.test.ts (CJK wrap content preservation, continuation indent, full-band highlight via exact ansi.reverse equality so it is color-level independent, SGR re-open across wrap points, ellipsis clamp, and render()-level budget assembly) and the 175 pre-existing pi-tui-runner.test.ts screen tests.
• tsc -p packages/cli/tsconfig.json clean; biome check clean on changed files.
• Manual: rendered the screenshot's long CJK options at width 40 — full text survives across wrapped rows (test pins this).

Review focus

Two rounds of adversarial sub-agent review were run against this diff. First pass found the missing height budget (major) and colorless-CI-vacuous assertions (minor), both fixed. Second pass found the title-cap ellipsis was dead code on already-padded lines (fixed via clampRowsWithEllipsis) and confirmed the remaining edge: a terminal too short for one row per option still overflows and falls back to the pre-existing pi-tui clip — never worse than before.

AI use

• Generative tooling made a substantive contribution

Tool(s) and scope: Maka (AI agent) authored the full change: issue triage, design, implementation, tests, and two adversarial review rounds (also AI, with human-supervised fixes).

Checklist

• Tests cover the change and fail without it
• Lint, format, typecheck and the affected suites pass locally

Does this PR entail a change in behavior?

• Yes — described under Summary above

[hqhq1025]

Found one correctness issue in the resize path.

Automated review notice: This comment was posted by an automated review agent operated by hqhq1025. It is not an independent human review and does not replace one.

[M4n5ter]

Reviewed exact head f00cdc58e2620137bda32939ae1cc0f6346aa55a.

The resize defect is closed. The bottom overlay now uses a percentage cap that the compositor resolves against the current terminal height, while the question surface reads its row budget on every render. The small-to-large terminal regression exercises the previously failing path and preserves both the free-text row and the full option text after growth.

I found no blocking or non-blocking correctness findings on this head. The exact-head test check is green and the current-main merge result is clean.

---

Automated review notice: This comment was posted by an automated review agent operated by M4n5ter. It is not an independent human review and does not replace one.