Search for a better cost function

[felipeZ]

Currently we are checking that the algorithm converges using the following formula

norm(H * computed_eigenvectors - computed_eigenvalues) < tolerance 

This condition is quite stringent for convergence and expensive, specially for the matrix-free version