fix: mirror explicit left/right paragraph alignment for rtl

[artem-harbour]

Linear: SD-3093

Fixes RTL paragraph alignment rendering parity with Word in pm-adapter.

What changed

• Updated paragraph alignment normalization so explicit left/right are mirrored for RTL paragraphs (w:bidi context).
• Added/updated regression tests for RTL alignment mapping and bidi fallback behavior.

Why

For RTL documents with explicit w:jc values (notably w:jc="right"), SuperDoc could render alignment opposite to Word. This patch aligns rendering behavior with OOXML/Word semantics.

Scope

Render pipeline only (layout-engine/pm-adapter), no toolbar/editor command changes.

[codecov-commenter]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[linear]

SD-3093

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: e56997752c

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve visual align commands for RTL paragraphs

When an existing RTL paragraph is edited through the TextAlign extension, setTextAlign('left'/'right') stores the requested visual value directly in paragraphProperties.justification (packages/super-editor/src/editors/v1/extensions/text-align/text-align.js:46). On the next render this normalizer now mirrors that PM attr, so clicking Align Left on an RTL paragraph writes left but DomPainter receives right (and vice versa). This makes the editor toolbar/shortcuts inverted for RTL paragraphs unless the command layer is updated to store the mirrored OOXML value or this mirroring is limited to imported OOXML semantics.

Useful? React with 👍 / 👎.