IRCAM database splitted into raw hrirs, diffuse field compensated hrirs, and max amplitude compensated hrirs.

signal is now 3 dimensional instead of 1 dimensional. Shape = (list of sounds, left/right ears, samples). This allows to feed the model with an arbitrary number of sounds, and allows to zero-pad the sound of each ear without affecting the ITD.

Added R-script to plot subjects performance by counting ratio of absolute angular errors less then X degrees (X in [5,10,15]).