fix problem in implementation of least squares fractional delay filter

[fietew]

If a fractional_delay of 0.0 is given to function, the former implementation leads to NaNs. Instead MATLAB's sinc function is used.