AbstractVoice v0.9.0

Added

• First-class lightweight install profile: pip install abstractvoice now keeps remote OpenAI-compatible audio adapters, AbstractCore plugin registration, shared voice types, and dependency diagnostics importable without installing local speech runtimes.
• Explicit install extras for deployment intent: abstractvoice[voice], abstractvoice[piper], abstractvoice[stt], abstractvoice[audio-io], abstractvoice[openai], abstractvoice[openai-compatible], abstractvoice[remote], and abstractvoice[all].
• Import-boundary tests that verify base package imports, plugin entry point discovery, and remote VoiceManager construction do not require Piper, faster-whisper, PortAudio/audio-device packages, or local audio file libraries.

Changed

• Moved local Piper, faster-whisper, microphone/playback, and audio-file runtime dependencies out of the base package and into explicit extras.
• Kept abstractvoice[all] as the common local feature bundle while leaving GPU-heavy Chroma, AudioDiT, and OmniVoice behind their own explicit extras.
• Made abstractvoice.audio and abstractvoice.adapters import local/audio dependencies lazily so remote and plugin deployments stay lightweight.
• Updated dependency diagnostics and user-facing install hints to distinguish lightweight base, local TTS, local STT, and audio I/O requirements.
• Refreshed README, installation, getting-started, dependencies, FAQ, API, REPL, and backlog docs for the lightweight base install contract.