Source code clarity of `recurrence_matrix`

At the moment the source code of recurrence_matrix, which is where the central computations of a recurrence matrix take place, is full of duplication. There version complexity scales as 2 x 2 x 2 = 8, with a multiplicity of 2 for (recurrences vs cross-recurence), (parallel vs non parallel), (dataset vs vector).

I do not believe that having 8 different methods is necessary. In fact, it may even be that we could do everything with 1 or 2 methods, by defining smaller helper functions.

In any case, a multiplicity of 8 is certainly not necessary and the source code can be simplified to remove duplication and thus make long term maintanance also easier.