Trtopi performance

Improved Backbone performance significantly when using large amounts of time steps in a single simulation. Tested to be working with VaGe Nordic data 24h simulation and a full year building model run with identical results, but I'm still unsure if these changes work with the long-term stochastics.

The performance of generating the limits of the reserve variables, as well as the performance of some of the larger equations (q_balance, q_obj) could potentially still be improved.