Native voice moves into the core

Slida 0.2.4 is coming with voice. Native whisper.cpp and ONNX inference support now live in the core, giving Slida a local path from spoken input to useful, structured work.

This release is the integration milestone, not the end of the performance work. Speed and efficiency will vary across the CPUs, GPUs, and accelerators people actually use while we tune the stack for more devices.

One day, we’ll get it optimized for all this hardware. Testers can now yell at AI.