feat(tts): add modern OpenAI audio providers

[RayJiang4S]

Summary

• add openai_audio_tts for audio-output chat models such as gpt-audio-1.5
• refresh openai_tts to use the current Speech API REST surface and updated voice/model options
• document OpenAI provider options and keep tts_selector ranking constrained by allowed providers

Tests

• python -m pytest -q tests/tools/test_openai_tts.py tests/tools/test_openai_audio_tts.py tests/contracts/test_phase3_contracts.py

[RayJiang4S]

Updated this PR with a routing fix from the recent multi-provider TTS audition workflow.

What changed:

• Added preferred_tool to tts_selector so callers can explicitly request openai_audio_tts instead of relying only on provider-level matching.
• Added prefer_audio_output so preferred_provider: \"openai\" can intentionally route to the openai_audio chat-audio path.
• Recorded instructions_applied and strict_script in openai_audio_tts output metadata for review/audit pages.
• Documented the selector routing knobs in docs/PROVIDERS.md.

Validation:

• .venv/bin/python -m pytest tests/tools/test_openai_tts.py tests/tools/test_openai_audio_tts.py -q passed: 8 tests.
• .venv/bin/python -m pytest tests/contracts/test_phase3_contracts.py -q passed: 63 tests.
• .venv/bin/python -m py_compile tools/audio/openai_tts.py tools/audio/openai_audio_tts.py tools/audio/tts_selector.py tests/tools/test_openai_tts.py tests/tools/test_openai_audio_tts.py passed.
• git diff --check passed.