Beam's alpha_i value can also be given as positive value now (internally it needs to be <=0 for incoming beams)

Refactored OpticalFresnel and removed inconsistency there (t should be defined such as to preserve energy); added basic stuff to enable roughness in matrix formalism

  a result; ideally it should not at all change the internal state;
  definitely it should not store the result.
Comments, layout.

This reverts commit e202d1ef.

Status: FitTest02 fails.

- speed as before
- App --isgisaxs07 shows huge difference, but fu.test passes