Fft odd results on windows(ooura)

[q-depot]

Hi,

there is an issue with the fft on windows, the first bin is always 0 and the last bin looks like the first one, so this could be related to a wrong index in the fft results in libXtract.
Lots of other features don't seem to be working, but I guess it depends on the wrong results generated by the fft.

[iver56]

My libXtract binaries for Windows (run as a vamp plugin with sonic annotator as host) seem to be producing wrong results for spectral centroid and bark bands. I guess the broken FFT in the build for Windows is the culprit.