Introduce native CSS-style Grid for the editor layout

[Michael Greene (mgreenegit)]

Summary

Introduce a dependency-free, CSS-style Grid primitive and use it for the viewport/editor layout, replacing the editor's manual height subtraction. This is a first replacement slice, not an enhancement to the legacy table engine.

This pull request was drafted by a large language model (LLM).

Scope

• Explicit display: grid, independent row/column templates, and flat row-major child placement.
• Intrinsic Auto tracks, fixed terminal-cell sizes, and integer fractional tracks. Fraction(n) means minmax(0, nfr), so editor content can shrink inside its viewport.
• One allocator for both axes. Auto tracks share unused space when no positive fractions exist; overflow is clipped.
• The existing implicit root fills the viewport. The editor reserves intrinsic space for chrome and Search/Replace, leaving the rest to the document.
• No dependencies, production-code relocation, or console/provider changes.

The equal-pane declaration is:

ctx.attr_display(Display::Grid);
ctx.attr_grid_template_columns(&[GridTrack::Fraction(1)]);
ctx.attr_grid_template_rows(&[
    GridTrack::Fixed(1),
    GridTrack::Fraction(1),
    GridTrack::Fraction(1),
    GridTrack::Fixed(1),
]);

Templates are typed Rust, not parsed stylesheets. Intrinsic measurement remains Edit's existing model, not browser min/max-content sizing. This does not claim full CSS Grid compliance.

Migration boundary

Legacy tables, their APIs, and menu/dialog declarations remain unchanged. Their layout path temporarily coexists with Grid; this PR does not claim to remove it. Shared collapsed-rectangle handling keeps zero-sized grid descendants valid.

The follow-up #961 unifies the engine using boxed column subgrids with preserved row focus, highlighting, and hitboxes. #962 then migrates the remaining callers and removes the table API without changing that shared engine. These interaction and migration contracts are intentionally outside this first main-view slice.

Review/merge order: #950 -> #961 -> #962. Both follow-ups link exact incremental comparisons because their full Files changed tabs include the unmerged prerequisites.

Each PR contains one commit. The follow-ups link content-identical review snapshots to isolate their incremental scope.

Review and validation

Five files: 254 changed production/documentation/registration lines, plus 343 test lines (597 total, down from 604). A private item-placement helper adds three lines versus the previous revision but reduces replacement in #961; the TUI-only pass removes 16 incremental changed lines across the series. Collapse/override fixtures share setup; alignment cases reuse a TUI.

Latest local validation: six Grid tests and two application layout tests pass, along with formatting, pre-commit workspace Clippy, and scoped structural checks. Test files are unchanged by the TUI-only pass.

Final-head Windows/Ubuntu CI and CodeQL pass at a23b7ac.

Regressions cover both-axis placement, intrinsic and fractional sizing, rounding, nested/padded and collapsed grids, floats, constrained alignment, repeated viewport resizing, editor typing/scrolling/search focus, and legacy menu highlighting and trailing-edge mouse activation under a grid ancestor. The shared interaction scenario checks typing after resize, closing Search, and closing a menu, for both empty and long documents.

An independent read-only review found no concrete correctness blocker. Its requested grid/legacy-menu interaction regression and subset documentation are included.

[Leonard Hecker (lhecker)]

I don't believe this is the right direction. My intention was to implement CSS grid & flexbox support in the near term (or at least a subset of those). Like this: https://jsfiddle.net/vkubqc4d/