DO NOT MERGE: Match pixels, but offset in time (for ex-ante forecasting of additionality)

[abbyevewilliams]

The idea for this very experimental branch is to create a set of matched pixels which represents the project, but offset in time. This is useful for making ex-ante predictions of deforestation for projects which haven't yet started.

I've tried to match the deforestation characteristics (LUCs and CPCs) of the project at t0 to those of the matched pixels at t-10. So if a project is 95% forested at t0 (now), we create a set of matched pixels which was 95% forested at t-10 (10 years ago). We can then measure deforestation in the matched pixels in the past 10 years to get a sense of what the next 10 years are likely to be like for the project. I've done some slightly janky renaming of columns in order to achieve this.

Here's a sketch which hopefully makes more sense:

The additionality output won't mean anything here but the resulting pairs are quite useful for further analyses.

In keeping with the format of the branch aew85-evaluations I have added a new folder, evaluations, which contains ex_ante_evaluation_template.Rmd as well as the scripts called by the R markdown file and a methods diagram which gets knitted into the final output. scripts/tmfpython.sh has been updated to match. This allows you to run the whole branch and knit an ex-ante report in one go.

--- There are many, many places where the code could be improved!!! ---