Multiple convergence-tolerances for gpe-diagonalization
Right now we pass in one value that has wildly different interpretations. For groundstates, it relates to either the differences between states in consecutive iterations (mixing), or the variance of the energy (optimal damping). For higher levels, it again relates in some way to differences between states, but differently than for gs-mixing. We should probably have at least two values, one for the gs and one for excited states, maybe even a list of values, one for each calculated state.