fix: zoom tears on a tall scroll region; terminal cursor follows the zoomed program

[cmj0121]

What

Three fixes to the zoom, from investigating #139 and #140.

• Clamp DECSTBM to the emulator's height (internal/vtirm). The daemon log holds 539 recovered an emulator write panic entries since June, every one in ultraviolet.(*Buffer).DeleteLineArea via vt.(*Screen).DeleteLine. Reproduced: a CSI 1;50 r on a 24-row emulator leaves the scroll region larger than the buffer, and the next DL / IL / RI / SU / SD indexes past it. The recover drops the rest of the chunk, so the panel shows a torn screen until the next full repaint. A program asks for such a margin whenever it still believes an older, taller PTY size, the window every zoom passes through between its attach and its resize. Upstream (x/vt 2026-09-27, ultraviolet 2026-09-29) does not clamp either, so the filter that already rewrites IRM now clamps margins on the way in, as a real terminal does.
• The terminal cursor follows the zoomed program (tea.View.Cursor). The zoom painted a reverse-video cell and left the real cursor at the end of the footer, which is where an input method draws the text it is composing: CJK marked text there wraps the line and scrolls the alternate screen out from under the diff renderer, for claude and grok alike. Verified end to end: with a real daemon on a private socket and a real grok panel, the terminal cursor now sits on grok's prompt cell (and advances four cells for 中文) where it used to sit on the footer.
• clipVisible clips by cell. A scrollback line of wide glyphs captured at a wider size was clipped by rune count and could overflow its tile by up to twice its width.

Verified not the cause

The emulator renders grok's frames (truecolor per-cell backgrounds, ?2026 synchronized output, SGR 59, braille art) identically to a reference at 80x24 and after a 120x40 resize; the replay+resize path from #133 reproduces the reference from any ring cut; an end-to-end cockpit run with a real grok panel rendered correctly before the change.

Tests

• vtirm: table of DECSTBM rewrites, no clamp without Rows, a sequence split across chunks.
• tui: every line op after a too-tall margin keeps the bytes that follow it; an in-range region still scrolls as asked; View().Cursor follows the emulator (wide glyphs included) and is withheld in history, selection, input overlays, the dashboard, and a too-small cockpit; clipVisible on CJK; golden zoom frames regenerated (only the painted cursor cell changed).
• make ci green locally.

Deferred

Interact mode in a group split still composes wherever the renderer left the cursor; several tiles share one terminal cursor, so that needs its own design.

Closes #139
Closes #140

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 92.59259% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 88.36%. Comparing base (e3d66dc) to head (0417daa).

Files with missing lines	Patch %	Lines
internal/tui/tui.go	81.81%	1 Missing and 1 partial ⚠️
internal/vtirm/vtirm.go	94.59%	1 Missing and 1 partial ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #141   +/-   ##
=======================================
  Coverage   88.35%   88.36%           
=======================================
  Files         149      149           
  Lines       23219    23266   +47     
=======================================
+ Hits        20516    20559   +43     
- Misses       1912     1913    +1     
- Partials      791      794    +3     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.