fix: preserve zero-width graphemes in the cell round-trip

[alexispurslane]

Problem

printString (styled.go) drops zero-width graphemes when decomposing a styled string into cells. Width-0 sequences (ZWNJ U+200C, ZWSP U+200B, combining marks that decode as width-0) are accumulated into the pending cell in the default branch (cell.Content += string(seq)), but the next visible cell overwrites them:

case 1, 2, 3, 4:
    cell.Width = width
    cell.Content = string(seq)  // clobbers the accumulated zero-width graphemes

So any content containing a standalone width-0 grapheme loses it on the round-trip. A decomposed accent renders wrong: "e\u0301x" comes back as "ex".

Repro

ss := uv.NewStyledString("e\u0301x")
area := ss.Bounds()
buf := uv.NewScreenBuffer(area.Dx(), area.Dy())
ss.Draw(buf, area)
// buf.Lines[0] is [ {e}, {x} ] — the combining acute is gone

Fix

Prepend the accumulated zero-width content to the next visible cell instead of overwriting it. Input order is preserved (the accumulated sequences precede the visible grapheme in the input), cell widths are unchanged, and the end-of-string flush already emits a trailing width-0 cell, so trailing graphemes survive too.

Test

Two cases added to TestStyledString: ZWSP between two letters and a combining mark before a letter. Both fail on main and pass with the fix.

Context

Found while building a TUI editor: the canvas round-trip (lipgloss layer compositing) silently stripped ZWNJ and combining marks from editor content every frame. Note that #95 is adjacent (unknown escape sequences are dropped in the same default branch) but is a separate bug.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 60.52%. Comparing base (19049f2) to head (c9372ef).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #151      +/-   ##
==========================================
+ Coverage   60.49%   60.52%   +0.02%     
==========================================
  Files          52       52              
  Lines        6751     6751              
==========================================
+ Hits         4084     4086       +2     
+ Misses       2391     2389       -2     
  Partials      276      276              

☔ 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.

[alexispurslane]

CI note: the conformance failures are pre-existing on main

The red conformance job is not caused by this change. I diffed the failing seeds:

Failing test	main nightly (run 31243567629, 06:17 today)	this PR (run 31256704853)
FuzzRenderer/8e096e0ae397d854	fails	fails
FuzzScreenShowsContent/seed#28-31,34-37	fails	fails
FuzzRedrawResyncs	fails (new crasher from tonight's fuzz job)	n/a (fuzz job is schedule-only)

PR-only failures: none. Every failing seed in this PR's run fails identically on main's own nightly test job, which ran hours before this PR existed. The committed crashers in internal/conformance/testdata/fuzz/ fail on every pull_request until the renderer bugs behind them are fixed — related territory to #140/#148/#150 (wide-cell handling across resizes).

The root-module CI (unit tests, go test ./... on main + this branch) is green; the new TestStyledString cases are red without the fix and green with it.

[alexispurslane]

My agent commented the solution before even asking me, sorry about that! But yes, this appears to be a pre-existing issue.

[taciturnaxolotl]

if you rebase on main the conformance test should be clean now!