Audio tests

[casey]

We need an easy way of testing audio. One idea:

• Figure out how to inject audio data into the analyzer node that is perfectly predictable, perhaps because the period exactly matches the buffer sizes, so you always capture the same time domain data, and thus the FFT always gets the same frequency data.

Once we have this, we can write tests against the audio reactive masks, as well as check that set_decibel_range does the right thing.