Speech context and device capture

Now you can pass in an array containing words/phrases so that the speech recognizer is more likely to recognize them!

You can also specify an input device when using arecord.

For usage info see: docs/API.md