Develop pointwise Riemann solvers about riemann

[mandli]

It would be nice if in the transverse point-wise solvers we included not only the row above and below of the aux array (as is done now) but also of the q array. This would not be backwards compatible with the implementation suggested as only a row of q is passed to the rpt routines.

[mandli]

Another issue with using rp* as the wrapper is that to get the interface above to work a temporary array must be created to store the correct values of aux1, aux2 and aux3 in aux_l and aux_r. This could be avoided again if the point-wise solver was called one level up in for instance flux2.

[mandli]

May want to let the window of the cells passed to both the normal and transverse solvers (not just the local ones).

[ketch]

Closed via #91.