It use our standard resampler to do it and so we do not keep
pitch yet (I am working on a soundtouch resampler).
Please test any regression.