Add gpu Cuda Sparse Matrix compatibility to function bicgstabl

[davidschlegel]

Although the conjugate gradient method cg does succesfully implement Cuda compatibility, bicgstabl does not.
I think it would be fairly simple to add this functionality.
Currently, using Cuda Arrays in bicgstabl is using scalar operations on GPU arrays, which should not be the case to have a performance speedup when using gpus.
I think an implementation by someone with a little bit of knowledge of CUDA should be straightforward.