CobbleVox: Voice Commands for Cobblemon

CobbleVox 1.0.2

- The speech engine (whisper.cpp DLLs) and the voice model (small-q5_1) are now bundled inside the mod. Nothing is downloaded at runtime, and no executable files are included.
- Recognition now runs inside the game through JNA (which ships with Minecraft) instead of a separate process, so no helper process stays running after the game closes.
- Faster recognition: flash attention is enabled, the default audio context is now 512, and the default thread count is half your CPU cores (up to 8). Each phrase takes about half the time it did before.
- Voice loads automatically when you join a world. No setup needed.
- Voice Manager simplified: the model install, reinstall and delete options are gone, since everything comes with the mod. Mode, Language and Microphone are still there.
- Removed the model and useServer config options, which no longer apply.