Initial release.

Extracted from SNGNav's voice_guidance package (https://github.com/aki1770-del/sngnav), where the LinuxTtsEngine drives in-vehicle voice guidance in production.

• TtsEngine abstraction (isAvailable, setLanguage, setVolume, setSpeechRate, speak, stop, dispose).
• LinuxTtsEngine: spd-say / speech-dispatcher backed, offline / on-device. Injectable process hooks for unit testing without a running daemon.
• NoOpTtsEngine: silent fallback for tests, headless, and non-Linux targets.
• Conditional export so the LinuxTtsEngine API is available on every platform (degrading to a silent NoOp where dart:io is unavailable).
• TtsUnavailableException + LinuxTtsEngine.ensureAvailable() for callers that require speech and must fail loudly rather than degrade silently.

This extraction excludes the Flutter-coupled FlutterTtsEngine from the source package, keeping linux_tts dependency-free (no flutter, no flutter_tts, no ffi) so it is usable in pure-Dart embedded contexts.