Use of astropy.convoultion instead (or optional) of scipy.signal routines

[Christian Herenz]

Astropy > 2.0 has a really good convolution implementation, that is better suited for astronomical data. It appears trivial to change to those routines, although we want to do also some tests.