General routine to interpolate (or even extrapolate small amounts) audio when a few of the samples are bad. Works great for a few dozen bad samples, but not so well with hundreds. Uses the least-squares autoregression (LSAR) algorithm, as described in:
Based on LoadLadspa, this code loads pluggable Audacity extension modules. It also has the code to (a) invoke a script server and (b) invoke a function returning a replacement window, i.e. an alternative to the usual interface, for Audacity.