Summary

• Keep a long internal pause when at least 1.5 seconds of speech remains after it.
• Continue cutting a long pause when the remaining audio is only a short noise tail.
• Cover long continuation, short tail noise, multiple pauses, and no-pause output with regression tests.

Verification

• Repository validation completed successfully.
• This validation record does not claim manual/browser verification.
• The real-browser E2E suite was not run.

Commands run:

./.venv/bin/python -m pytest backend/tests/test_audio_trim_continuation.py backend/tests/test_kokoro_trim.py -k 'not kokoro_backend_generate_applies_trimming' -q
./.venv/bin/python -m compileall -q backend
./.venv/bin/ruff check --config backend/pyproject.toml backend/tests/test_audio_trim_continuation.py
./.venv/bin/ruff format --check --config backend/pyproject.toml backend/tests/test_audio_trim_continuation.py

Fixes #1108