This fixes an issue with a-Amp interpolating the parameter in dB, resulting in a double-exponential fade when the parameter changes. Now fade is linear in dB, also using Ardour' Amp processor is more efficient, since interpolation happens in C++.
