Change linear AMR model to same form as logarithmic

[cgarling]

$\langle [\text{M} / \text{H}] \rangle(t) = \alpha \left( T_{max} - t \right) + \beta$ rather than $\langle [\text{M} / \text{H}] \rangle(t) = \alpha t + \beta$ as it currently is.

This will make presentation clearer in the paper as the logarithmic AMR is $\langle Z \rangle(t) = \alpha \left( T_{max} - t \right) + \beta$ so the two have a common form. This also allows us to enforce $\alpha>0$ which makes more sense as a slope than $\alpha<0$ as it was in the old model. It will necessitate some changes to scripts with science results to deal with the new formats but I think the change is worth it.

[cgarling]

Can convert the old $\alpha_1$ and $\beta_1$ to new ones $\alpha_2$ and $\beta_2$ as $\alpha_2=-\alpha_1$ and $\beta_2=\beta_1+\alpha_1*T_{max}$.

[cgarling]

With this change we are now mixing how we treat time bounds; some methods take max_logAge while others take T_max. We should make this uniformly T_max while making this change.