Grapheme-cluster aware cursor, hit-testing and soft wrap on top of #5922

[xyos]

Builds on #5922 (initial range.getClientRects()-based text measurement). This adds the two commits on top of that branch making the cursor/column model and soft wrap grapheme-cluster aware, and fixing related rendering issues found while testing against a real browser.

Changes on top of #5922

Grapheme-cluster cursor model & pixel-accurate coordinates (50aff8d)

• New lang.getGraphemeCluster() / lang.getGraphemeBoundaries() helpers using Intl.Segmenter (grapheme granularity), with a surrogate-pair fallback when unavailable.
• Selection.moveCursorTo snaps to grapheme cluster boundaries instead of only surrogate pairs, so arrow keys traverse ZWJ emoji (👨‍👩‍👧‍👦) and combining marks (ã̤) in one step.
• VirtualRenderer.textToScreenCoordinates uses fontMetrics.textWidth() instead of column × characterWidth, so forward and inverse coordinate mappings agree.
• FontMetrics.$pixelToColumn hit-testing iterates real grapheme clusters, treats span seams inclusively (fixes clicks with showInvisibles falling between spans), and clamps clicks left of line start to column 0.
• U+200D removed from the invalid-invisible regex so ZWJ emoji render intact (fixes Emojis consisting of surrogate pairs are displayed incorrectly #5813).
• Line-start RLE (U+202B) renders as plain text instead of a red invalid dot when rtl/rtlText is enabled (fixes ace.js: placing a red dot when clicking "enter" #5423, same approach as Fix: make RTL new line a valid character #5434); Trojan-Source highlighting still applies outside RTL mode.

Grapheme-aware soft wrap (5d2bc1d)

• $getDisplayTokens marks continuation code units of grapheme clusters with a new CHAR_CONT display token (only for lines containing chars ≥ U+0300).
• $computeWrapSplits never splits inside a cluster: a split landing mid-cluster moves back to the cluster start, or past the cluster when it is wider than the wrap limit. Before this, a wrap limit crossing 👨‍👩‍👧‍👦 split its 11 code units across screen rows.

Issues this addresses

#460, #4142, #5431, #5813, #4602, #5423, #5436, #3753, #3866, #3617

Testing

• experiments/widthchar.html — a browser harness with monospace/proportional/RTL/wrapped editors over pathological lines (BMP & ZWJ emoji, combining marks, CJK, tabs, Hebrew, U+FE0F, APL fallback glyphs). It verifies cursor DOM position against Range ground truth, textToScreen→screenToText round-trips at every grapheme boundary, arrow-walk stops, selection-marker coverage, wrap-split boundaries, and the RTL red-dot/backspace scenarios: 2,400+ assertions passing in Chrome, 0 failures (baseline before these commits: ~280 failures).
• New unit test wrapLine split never breaks grapheme clusters in edit_session_test.js; full suite unchanged otherwise (1,480 passing).

Open kitchen-sink @ 5d2bc1d6012061d91e50b3bf433c8ea27376b99a

Testing Page

Open kitchen-sink @ 423a67e091c51d742baaa72bd1d0a2aca284f72f

[github-actions]

One of the public type files has been updated, plase make sure there are no backwards incompatible changes done in the PR.

[codecov]

Codecov Report

❌ Patch coverage is 93.50153% with 85 lines in your changes missing coverage. Please review.
✅ Project coverage is 93.93%. Comparing base (c89c0d5) to head (423a67e).
⚠️ Report is 8 commits behind head on master.

Files with missing lines	Patch %	Lines
src/layer/font_metrics.js	93.88%	32 Missing ⚠️
src/test/mockdom.js	93.54%	18 Missing ⚠️
src/layer/text.js	61.29%	12 Missing ⚠️
src/edit_session.js	83.33%	10 Missing ⚠️
src/lib/lang.js	88.57%	8 Missing ⚠️
src/selection.js	55.55%	4 Missing ⚠️
src/layer/marker.js	93.33%	1 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #5988      +/-   ##
==========================================
+ Coverage   93.73%   93.93%   +0.19%     
==========================================
  Files         640      641       +1     
  Lines      137743   138724     +981     
  Branches    14487    14757     +270     
==========================================
+ Hits       129113   130306    +1193     
+ Misses       8630     8418     -212     

Flag	Coverage Δ
unittests	93.93% <93.50%> (+0.19%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ 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.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[xyos]

Pushed a third commit (423a67e) completing the column model rework: 1 grapheme cluster = 1 screen column.

Screen columns now count grapheme clusters instead of UTF-16 code units, addressing the "column must be column" model discussed in #5431:

• $getStringScreenWidth walks grapheme boundaries (both documentToScreenPosition and screenToDocumentPosition funnel through it), gated by a cheap regexp so plain text skips segmentation.
• $getDisplayTokens emits one display token per cluster with a parallel docLengths array; $computeWrapSplits consumes it, so wrap budget is now measured in visual columns — a line of ZWJ emoji no longer wraps 11× too early — while wrap split offsets remain code-unit doc columns.
• Text layer and font-metrics DOM walkers translate between grapheme screen columns and code-unit text-node offsets.

Verified in Chrome via the experiments/widthchar.html harness (2,800+ assertions incl. new graphemeColumns and vertical-navigation desired-column checks, 0 failures) and the node suite (1,480 passing; the 2 failures are pre-existing on master-adjacent branches).

[github-actions]

One of the public type files has been updated, plase make sure there are no backwards incompatible changes done in the PR.