Sparse matrices as entry for solvers

We expose our solvers (from `Eigen/IterativeLinearSolvers`) with dense matrices as entry (`MatrixXd`). 

Yet, in `Eigen`, they are exposed and tested with sparse matrices too (`Eigen::SparseMatrix`) - in solvers like `BiCGSTAB`, `MINRES`, `ConjugateGradient`, `LeastSquaresConjugateGradient`

Is it of interest to expose them with sparse matrices in `nanoeigenpy` ?