Better support for "adapt in Init()" AMI models.

This release improves support for "adapt in Init()" AMI models.
Previously, the Rx AMI model Init() function was being sent an ideal impulse, which meant that it couldn't adapt to the channel response. Now, that upstream channel response is being sent in and deconvolution is used to extract the Rx impulse response.

Also, the ability to save/load reference waveforms, for the impulse, step, pulse, and frequency responses, has been added. This facilitates easy comparison of Init() vs. GetWave() behavior.