Alterations for a more general-purpose design

Currently we use the lick detectors in an ephys context and are trying to use them in a more general purpose behavior context.

There are a few issues that prevent us from using these devices in general purpose cases:

* Two lick detectors of the same frequency interfere (unbounded excitation signal growth) with each other with the original AFE (i.e: 2nd order bandpass filters) when placed in close proximity to each other (< 5mm).
* Because the lick detectors are high impedance, connecting a plastic water tube to them can cause them to spuriously trip when the water tube makes intermittent connection to conductors, large metal things, etc.
* Two lick detectors in close proximity to each other (<5mm) of the same frequency can pickup licks from the other lick detector. I.e: Licks on one detector can cause a spurious noise artifacts (akin to a solenoid) emerging on the other lick detector. This is easier to filter out on the original AFE (2nd order bandpass filter) but more difficult to filter out on the new AFE without extending the detection window past the 1[ms] deadine.
* Lick detectors can be susceptible to spurious 60Hz noise after the start of the lick when the mouse is not earth grounded. The original AFE was susceptible to this issue (2nd order bandpass filter), but the current one (4th order bandpass filter) is not.

There are a few fixes that can make the lick detectors generally more useful in all cases:
* Make the series impedance resistor selectable. (100KOhm and 10KOhm would be two good choices)
* Make excitation voltage amplitude selectable. (20mVpp to 500mVpp)
* Provide lick detectors at two different frequencies (100KHz and 200KHz?), such that two lick detectors can be placed in close proximity to each other.
* Revert to a 2nd order bandpass filter to increase detection response time and remove "bouncy" effects of cascaded filters, but increase the cutoff as much as possible.
* Enable dynamic gain on the output stage and configurable excitation voltage on the input stage for a more rugged signal in behavior contexts.