Fix Command Mode clarification card layout for long questions and multiple options #33

Summary

In Command Mode, Lucid sometimes asks a clarification question before applying an edit. When the question is long or it returns several choices with longer labels, the clarification card becomes difficult to use: text and controls crowd or overlap, producing layered/illegible content.

Current UI path

The clarification state is rendered in apps/desktop/src/dictation/DictationBar.tsx using .dictation-card, .card-text, and .card-actions from apps/desktop/src/styles.css.

The card currently places the question and option actions in the same horizontal flex row. The actions wrap within a maximum width of 55%, but there is no responsive layout for long content or narrower windows.

Reproduction

1. Enter Command Mode.
2. Submit an ambiguous edit that causes the AI to ask a clarification question.
3. Use a case with a longer question and 3–4 options, preferably with multi-word option labels.
4. Observe the card at ordinary and narrower desktop window widths.

Actual behavior

• Question, helper text, and option buttons can crowd or overlap.
• Multiple wrapped controls are difficult to scan.
• The clarification UI can obscure itself or nearby content.

Expected behavior

The clarification card should grow and reflow predictably. The question, helper text, every option, and Cancel must remain readable and selectable without overlapping or clipping.

Acceptance criteria

• Long questions and the maximum supported 4 options remain legible at all supported desktop window sizes.
• Option labels wrap safely; controls never overlap the question, helper text, input area, or editor.
• The layout stacks vertically when horizontal space is insufficient.
• Keyboard focus, click targets, and the voice-selection hint remain usable.
• The result works in light and dark themes.
• Add a UI/E2E regression case with a deliberately long question and long option labels.
• Proposal and content-status cards do not regress.

Non-goal

Do not change when or why the model asks for clarification. This issue is limited to presentation and interaction stability.