Added warning about using a complex shift where eigenvalue should be directly...

Added warning about using a complex shift where eigenvalue should be directly evalauted from a Ritz evaluation. One value will be correct but we do not know in general which one!