Request: Add Eigsh-Cupy Mode

[cgbsu]

CuPy also has eigsh and it makes computation much faster, the code should be almost identical to what is alraedy there

        if method == 'eigsh':
            from scipy.sparse.linalg import eigsh

            # Note: uses shift-invert trick for stability finding low-lying states
            # Ref: https://docs.scipy.org/doc/scipy/reference/tutorial/arpack.html#shift-invert-mode

            eigenvalues, eigenvectors = eigsh(H, k=max_states, which='LM', sigma=min(0, self.E_min))

Simply change the import (and condition) and it should work.

[astralm0nke]

Added this and pushed in a new PR.