Slate v2 Instance Surface Recovery

Goal

Start True Slate RC tranche 2 by restoring an overrideable editor instance surface for the methods the current engine already supports.

Current Slice

1. add red tests for createEditor() instance methods and helper delegation
2. widen the Editor type and createEditor() instance shape
3. make Editor and Transforms namespace helpers delegate through instance methods where the engine already supports those seams
4. run targeted slate tests

Notes

• This slice is not full legacy parity.
• withoutNormalizing is still a separate explicit decision unless the current transaction seam proves it is an honest alias.

Result

• createEditor() now attaches an overrideable instance surface for the methods the current engine already supports.
• Editor.* delegates through that instance seam.
• Transforms.* stays low-level and unchanged, so the slice restores the override seam without silently changing transform semantics.
• the slate source barrel again exports the runtime helper surface currently consumed by sibling packages: Node, Text, Element, Range, Path, and isObject.
• the next behavior-method seam (deleteBackward, deleteForward, deleteFragment) is now also restored on the editor instance and on Editor.* using the current delete transform semantics.
• withoutNormalizing is now restored as an explicit compatibility alias over the current transaction boundary.
• insertBreak and insertSoftBreak are now recovered for the currently proved top-level block split seam, including edge-position always behavior at the start and end of a text node.
• editor.isInline, editor.isVoid, Editor.isInline, Editor.isVoid, and Editor.isEditor are now restored on the current query seam, with EditableTextBlocks using editor.isInline as the default runtime fallback when no explicit isInline prop is supplied.
• markableVoid is now restored on the current seam, and addMark / removeMark honor it for the text child inside a void element when the hook returns true.
• normalizeNode is now restored as a real outer-transaction multi-pass seam, and the forced-layout example/runtime proof uses that seam instead of the old subscribe-based workaround.
• verification:
  • yarn test:custom
  • yarn lint:typescript