[Scenario] Check and enable support for backcasting

[Martin-Jung]

The current scenario projections assume that projections are made either for contemporary or future conditions, but not past conditions (e.g. using 2000 estimates for projecting suitability for the year 1950).
Although the process should technically be the same assuming correct date handling, there are some practical challenges with this as for example dispersal constraints assume that a sequential iteration of dates (t_0, t_1, ...) rather than the inverse (t_-1, t_-2). Could likely be fixed by a sorting of dates of some sort, but needs to be tested out.