feat(tts): add Azure AI Speech provider

[RayJiang4S]

Summary

Adds an Azure AI Speech text-to-speech provider for OpenMontage.

Highlights:

• REST synthesis via Azure AI Speech with AZURE_SPEECH_KEY and AZURE_SPEECH_REGION
• operation=list_voices for voice catalog discovery
• SSML generation with voice, style, role, rate, pitch, volume, and sentence silence controls
• Custom endpoint and custom voice deployment support
• Optional SDK mode for word-boundary timing metadata when azure-cognitiveservices-speech is installed
• TTS selector compatibility through the existing capability/provider registry
• Documentation and provider contract coverage

Validation

Passed:

.venv/bin/python -m pytest -q tests/tools/test_azure_tts.py tests/contracts/test_phase3_contracts.py
# 71 passed

Also validated manually with a live Azure Speech resource:

• provider status: available
• operation=list_voices for zh-CN: 49 voices returned
• direct SDK synthesis succeeded
• SDK word-boundary metadata returned 13 boundaries / 11 words for a Chinese sample
• tts_selector routed preferred_provider=azure to azure_tts and generated audio

Full-suite note:

.venv/bin/python -m pytest -q

currently stops during collection in tests/qa/test_08_end_to_end.py because its fixture does not include the now-required render_runtime property for edit_decisions. That appears unrelated to this Azure provider change.

[RayJiang4S]

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

What changed:

• Added operation: \"preflight\" to report Azure credential/region status, SDK availability, word-boundary intent, and fallback behavior before batch generation.
• Added require_word_boundaries; when false, Azure now falls back to REST synthesis if enable_word_boundaries requested the optional SDK but the SDK is unavailable.
• REST fallback records word_boundary_fallback: \"rest_without_word_boundaries\", empty words/boundaries, and a warning in output metadata so review pages can explain why timing metadata is missing while still producing audio.
• Documented the fallback and preflight workflow in docs/PROVIDERS.md.

Validation:

• .venv/bin/python -m pytest tests/tools/test_azure_tts.py -q passed: 10 tests.
• .venv/bin/python -m pytest tests/contracts/test_phase3_contracts.py -q passed: 63 tests.
• .venv/bin/python -m py_compile tools/audio/azure_tts.py tests/tools/test_azure_tts.py passed.
• git diff --check passed.