Fixes for the last remaining leaks in LanguageModelGenerator Fix for the bug in which OpenEars could only use .DMP files that it created itself
- OpenEarsConfig.h now only controls which voices are included in your app so you can slim down your binary size. Everything else that used to be in OpenEarsConfig.h is now dynamically controlled at runtime.
- Runtime logging control. No more recompiling the framework to toggle logging on and off, just set it from your app. OpenEarsLogging is controlled globally and PocketsphinxController and LanguageModelGenerator have their own loggers.
- Runtime calibration length control. No more recompiling the framework to change calibration time, just set it from your app.
- n-best! You can now see all the top hypotheses for the same utterance and their scores. You can set how many you want to see.
- Testing tool: you can now run one round of recognition on an arbitrary wav file for your testing enjoyment.
- ps_process_raw has now been set back to no_search = FALSE since related issues appeared to have cleared up.