Frequency detect mode for input[1]

[trentgill]

Closes #277

DO NOT MERGE

This PR adds a new input mode freq short for frequency tracker. It functions the same as stream & volume modes, but returns values representing the frequency of an attached oscillation.

The detection algorithm is just looking for zero-crossings, so it will only work with raw oscillator sounds (though perhaps could be fun to explore with complex waves for harmonic-locking effects).

Working accurately up to about 1500Hz in my testing, with increasing jitter above this point.

//

I did notice that all the values were off by about 2%. After some investigation, it's not caused by the detection algorithm -- i tested sending LFOs from ASL, and they were also this same 2% off. It seems that the dsp block itself is running slow. I haven't investigated whether this is caused by variation between devices (it could be as I don't think our crystal is very accurate in an absolute sense - there is no dedicated audio crystal), or if the PLL that controls the audio-loop is incorrectly configured. Perhaps worth some investigation, but it shouldn't hold up this PR.

The good news is the frequency offset scales exponentially. To correct for it, one simply multiplies by 0.98 or so in frequency land. If converting to hertz, it just becomes an addition.

//

Before merging, i want to add a hztovolts function based on the same absolute freq range as just-friends / w/syn etc.
The reason i'm waiting is it can leverage the new justvolts function. Should just be:

function hztovolts(hz, ref)
  ref = ref or 261.63 -- optional. defaults to middle C
  return justvolts(hz/ref)
end

This is especially handy if you want to track a frequency input and control a module over ii, or even to take an input frequency and send it to an output as a voltage (so it can be doubled by another oscillator). This would be especially interesting if the output had a .scale applied, and in volts space, adding an offset becomes a direct transposition.

//

Together with the new calibration engine & public libs, we should make a norns-enabled script for volt-per-octave calibration of other modules.

[tehn]

looks great. impressed at the deep implementation, and seeing MIDI code made me smile