test(docx-core): characterize terminal paragraph-mark cleanup parity

[stevenobiajulu]

Characterize terminal paragraph-mark cleanup parity

Independent dynamic Opus 5 review of PR #980 at d943a53 found a low-severity, pre-existing difference between native primitives and comparison AST when resolving a paragraph-mark revision with no following paragraph sibling:

<!-- primitives -->
<w:pPr><w:jc w:val="right"/><w:rPr/></w:pPr>
<!-- comparison AST -->
<w:pPr><w:jc w:val="right"/></w:pPr>

The empty mark properties are schema-valid. The early-return marker cleanup in packages/docx-core/src/primitives/accept_changes.ts predates PR #980's shared formatting repair. Do not conflate XML canonical parity with a demonstrated text, numbering or reader corruption.

• Reproduce independently on main with synthetic terminal paragraph input for both Accept and Reject.
• Check table-cell terminal paragraphs and filtered single-author projections.
• Decide whether normalized structural equality should ignore empty rPr or both engines should remove it consistently.
• Add parity regression plus schema validation without changing paragraph-removal semantics.

No Word behavior has been verified. No private inputs are needed. Keep separate from #980, #891/#973 terminal empty-paragraph behavior and #979 note reader gaps.

Ref: #941, PR #980 (review finding F4).