Add JSON, SRT, and WebVTT ASR exports

[jmayank1511]

Summary

• add a normalized transcript model and JSON, SRT, and WebVTT exporters
• add --output-file and --output-format to streaming, offline, and realtime file ASR clients
• automatically request word timestamps for subtitle output and preserve segments, alternatives, confidence, speaker, and language metadata

Testing

• python3 -m pytest -q tests/unit/test_transcript.py tests/unit/test_realtime.py — 15 passed
• Ruff checks passed for all changed Python files
• source-tree CLI help smoke tests passed for all three modified clients
• full unit suite: 38 passed; one unrelated existing TTS mock assertion failed in tests/unit/test_tts.py::TestSpeechSynthesisService::test_synthesize_online