libsamplerate 0.1.2 didn't have the 32 bit <-> float conversion
routines.  Emulate them in case they aren't supported.