fix: export the loader's canonical form for nested lists

[LaatonWalaBhoot]

Closes #81

Defines the canonical nested-list form as what a load can represent, and makes the export emit it, so toJson() → load → toJson() is a fixed point in every nested state:

• listItem chains keep their nesting — the loader preserves relative depth through listItem content, so nothing changes for bullet/ordered chains (a guard test pins this, including indentation across reload).
• A task item's nested lists are exported as document-ordered sibling lists one level below the item — on load, anything nested inside a taskItem re-enters at exactly that level, so deeper chains built by live edits restructured on reload. The export now flattens the subtree to the loadable shape.
• A lone line-break piece no longer becomes an empty text node — a loaded empty nested item carries its break as a standalone piece; toInlineNode strips the break and emitted {"text":""}, which the next reload dropped. Filtered next to the existing decorator/empty-piece filters.

Tests: NestedListCanonicalFormTest — 2 cases fail on main, plus the listItem-chain guard. Full suite passes, JS/Wasm compile clean.

With this and #83, the #46 seeded sweep runs 300 seeds × 250 ops with zero crashes and zero export divergences.

[Copilot]

Pull request overview

Updates TextKit’s piece-table → JSON export to emit the loader’s canonical nested-list shape (so toJson() → load → toJson() becomes a fixed point for deeply nested list states), and adds regression coverage for the previously divergent cases.

Changes:

• Add NestedListCanonicalFormTest to pin fixed-point behavior for nested list exports (including the “empty text node” regression).
• In PieceTableConverter, flatten task-item nested list subtrees to a loadable canonical form and filter lone line-break pieces to avoid emitting {"text":""}.
• Introduce a helper to flatten a positional-paragraph subtree in document order.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
shared/src/commonTest/kotlin/com/jjrodcast/textkit/NestedListCanonicalFormTest.kt	Adds tests asserting fixed-point export/load behavior for canonical nested-list forms.
shared/src/commonMain/kotlin/com/jjrodcast/textkit/editor/core/converters/PieceTableConverter.kt	Adjusts task-item nested-list export canonicalization and prevents emitting empty text nodes from lone line-break pieces.

Comments suppressed due to low confidence (1)

shared/src/commonMain/kotlin/com/jjrodcast/textkit/editor/core/converters/PieceTableConverter.kt:504

• nestedElements is built by calling createParagraphModel(listOf(positionalParagraph), ...) for every (flattened) nested list item. If a task item contains a nested list with multiple items, this will emit multiple adjacent list nodes (each with a single item) instead of one list node containing all items. This is inconsistent with how top-level/conventional list groups are exported (grouped into one list node per run) and can bloat/fragment the exported JSON for downstream consumers.

                val nestedElements = positionalParagraphs.flatMap { it.flattenSubtree() }.map { positionalParagraph ->
                    // endIndex is exclusive: (0, 0) sublists to nothing, so a nested list rebuilt
                    // here came out empty — `{"content":[],"type":"taskList"}` in the export, which
                    // a reload drops (issue #79). The single element needs (0, 1).
                    createParagraphModel(listOf(positionalParagraph), startIndex = 0, endIndex = 1)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[LaatonWalaBhoot]

@jjrodcast I think it can be merged