fix(elevenlabs): restore chunk_length_schedule in WS init payload

[IanSteno]

Summary

• restore generation_config.chunk_length_schedule in ElevenLabs multi-context websocket init packets
• keep existing default auto_mode=True for normal usage, but default to auto_mode=False when chunk_length_schedule is explicitly provided so the schedule is not implicitly disabled
• refactor init packet construction into a helper for deterministic unit coverage

Why

chunk_length_schedule has been accepted/stored by the plugin but not forwarded in current multi-context websocket flow, making it a no-op in practice.

Changes

• livekit/plugins/elevenlabs/tts.py
  • add _build_context_init_packet(...) to include:
    • text
    • voice_settings
    • context_id
    • generation_config.chunk_length_schedule when provided
    • pronunciation_dictionary_locators when provided
  • use this helper in _Connection._send_loop() init path
  • set constructor defaulting to auto_mode = not is_given(chunk_length_schedule) when auto_mode is not explicitly provided

Tests

Added tests/test_plugin_elevenlabs_tts.py covering:

• auto_mode defaults with and without explicit chunk schedule
• explicit auto_mode override precedence
• init packet includes generation_config when schedule is set
• init packet omits generation_config when schedule is unset
• pronunciation dictionary locators are preserved in init packet

Run locally:

• uv run pytest tests/test_plugin_elevenlabs_tts.py -q
• uv run ruff check livekit-plugins/livekit-plugins-elevenlabs/livekit/plugins/elevenlabs/tts.py tests/test_plugin_elevenlabs_tts.py

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 3 additional findings.