Allow ncore to be an iteration variable that calculates dene

[chris-ashe]

Following on from #736.

With errors seen in calculating $\mathtt{ncore}$ due to VMCON iterating $\mathtt{dene}$ & $\mathtt{neped,nesep}$ that can lead to negative $\mathtt{ncore}$ values as per:

$$\begin{aligned} \nonumber n_0 & = & \frac{1}{3\rho_{ped,n}^2} \left[3\langle n\rangle (1+\alpha_n) + n_{sep} (1+\alpha_n) (-2 + \rho_{ped,n} + \rho_{ped,n}^2) \right.\\ & & \left. - n_{ped}\left( (1 + \alpha_n)(1+ \rho_{ped,n}) + (\alpha_n -2) \rho_{ped,n}^2 \right) \right] \end{aligned}$$

This is mainly seen if the lower bound of $\mathtt{dene}$ is set too low and the upper bound of $\mathtt{neped,nesep}$ is set too high. This can easily be the case with a user with not much experience who thinks widening the bounds will make it easier to solve for VMCON.
When iterating VMCON can easily get into a point where the calculated $\mathtt{ncore}$ is negative.

A possible long term code solution is to create $\mathtt{ncore}$ as an iteration variable and re-arrange the equation above to calculate $\mathtt{dene}, (\langle n\rangle)$. This would make setting lower bounds easier as a >0 limit could be enforced. Though it would not be possible to have $\mathtt{ncore}$ and $\mathtt{dene}$ as an iteration variable at the same time .

@mkovari @timothy-nunn