Support for "sigma" in the NEWARP version of eigs_sym()

This patch adds the shift-and-invert mode for eigs_sym(), using the built-in NEWARP implementation. It uses SuperLU to compute the sparse LU factorization, similar to the sp_auxlib::eigs_sym_arpack() function.

I haven't done the eigs_gen() function, since I feel it may be more helpful to first receive some feedback regarding the change. @matteosecli @rcurtin @conradsnicta