fix: keep page text for bookmarked Flash summaries

[Iams4kura]

Summary

• Fixes: PageIndex Flash raises IndexError when a sparse PDF reaches the early outline gate, embedded bookmarks supply a non-empty tree, and summaries are enabled.
• Root cause: The early-return path passed extracted page text to bookmark merging but omitted it from the result, so the public API gave summarize_tree an empty page list for the bookmark-derived nodes.

Regression evidence

• Before: PYTHONPATH=. uv run --with-requirements requirements.txt --with pytest pytest -q -p no:cacheprovider tests/test_flash_api.py exited 1

• After: PYTHONPATH=. uv run --with-requirements requirements.txt --with pytest pytest -q -p no:cacheprovider tests/test_flash_api.py exited 0

Verification

• PYTHONPATH=. uv run --with-requirements requirements.txt --with pytest pytest -q -p no:cacheprovider
• PYTHONPATH=. uv run --with-requirements requirements.txt python -m compileall pageindex tests

Scope

• 2 files changed, +39 / -3 lines

Authorship

This fix was developed with AI assistance and was validated with the regression and project checks listed above.