This release packs a lot of refinement.
* The most important change is the model that estimates masking thresholds according to Johnston acoustic model. This means a much more intelligent adaptation of the reduction in order to avoid artifacts and distortion.
* The post-filtering option was eliminated in favor of the masking model because it not only reduce artifacts at low SNR levels but in all conditions and leads to a cleaner reduction.
* Other change that worth mentioning is the addition of a simple onset detector based on spectral flux and moving average thresholding to adapt the reduction. This is pretty useful to avoid softening transients in percussive material like drums or picked guitars.
* Whitening was revised too and now is adaptive to changing background noise.
* Code has been documented using doxygen
* Lots of other bug fixes and code refinements were introduced too.
A big thanks to everyone that helped directly and indirectly with this release!
