diff audioio/AudioGenerator.cpp @ 144:675443abbc05

Various fixes:
 * Fix handling of HTTP redirects (avoiding crashes... I hope)
 * Fix failure to delete FFT models when a feature extraction model
   transformer was abandoned (also a cause of crashes in the past)
 * Fix deadlock when said transform was abandoned before its source
   model was ready because the session was being cleared (and so the
   source model would never be ready)