v0.4.0

Added

• Native inference with model reuse between requests.
• configureWhisper(), getWhisperStatus(), and disposeWhisper() to manage concurrency, queued requests, and model cleanup.

Changed

• nodewhisper() prefers the native backend, with CLI fallback when the addon is unavailable or an option requires it.
• By default, one transcription runs at a time and up to 100 requests can wait. Additional requests reject when the queue is full; limits are configurable.

Fixed

• Support Pino and other loggers that provide info() without log().
• Native VAD respects the configured thread count, including when reusing a model, avoiding slowdowns on machines with fewer CPU cores.