Inconsistent energy in maxcut_qaoa depending on samples argument

[mrosenkranz]

The energy output during the optimisation in maxcut_qaoa.get_angles() (the line E => ...) approaches very different values for samples=None or setting a number of samples (e.g. samples=1000). The displayed value with samples=None doesn't seem to be the true expectation value of the cost Hamiltonian, whereas it matches the expected cost Hamiltonian when specifying a number of samples .