fix(webvtt): keep cue text in order after a multi-line span

[businessarshgoyal]

WebVTTDocumentBackend cached the current cue paragraph in a local variable before walking a cue payload. A nested span (voice, bold, italic, underline) that wraps a line terminator starts a new paragraph inside the recursive call, so after returning the caller was still appending to the paragraph it had cached — text following the span landed on the first line of the cue instead of the line it actually appears on.

For

00:00:01.000 --> 00:00:05.000
<v Bob>Hello
there</v> and afterwards

the cue used to produce ['Hello', ' and afterwards', 'there']; it now produces ['Hello', 'there', ' and afterwards'].

The fix appends to cue_text[-1] instead of a cached reference, which is by construction the paragraph the terminator handling last opened, at any nesting depth. No existing ground truth changes; the four WebVTT reference documents are unaffected.

Issue resolved by this Pull Request:
Resolves #4104

Checklist:

• Documentation has been updated, if necessary. (not necessary — no user-facing API change)
• Examples have been added, if necessary. (not necessary)
• Tests have been added, if necessary.

Verified with uv run pytest tests/test_backend_vtt.py (12 passed) and make check (ruff clean; the ty and max-lines findings it reports are pre-existing and untouched by this change). The new test fails on main and passes with the fix.

AI disclosure: the patch and test were written with an AI coding agent (Devin) and reviewed and run locally by me.

[github-actions]

✅ DCO Check Passed

Thanks @businessarshgoyal, all your commits are properly signed off. 🎉

[mergify]

Merge Protections

🟢 Merge protection satisfied — ready to merge.

Show 1 satisfied protection

🟢 Enforce conventional commit

Make sure that we follow https://www.conventionalcommits.org/en/v1.0.0/

• title ~= ^(fix|feat|docs|style|refactor|perf|test|build|ci|chore|revert)(?:\(.+\))?(!)?:

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[dolfim-ibm]

lgtm