make the output of ft_dipolefitting (more) similar to the output of ft_sourceanalysis

The `ft_datatype_dip` function has been empty since 2011, i.e. the structure that describes dipoles is still not properly defined. The only thing that `ft_datatype` knows about it is that it contains a `dip` sub-field.

This relates to #1399 where it is unclear how the units are to be resolved.

The result of a dipole fit (first gridsearch, then nonlinear optimization) is technically not different from the result of a MNE estimate on a single (or few) positions. Also, it would be nice if you could for example first do a dipolefit, and then feed the result in `cfg.sourcemodel` for a subsequent beamformer estimate.