Analysis to check agriculture and incorporate GDDs and KDDs

[jrising]

I think this is eventually going to include three things:

• The simulated-to-reality agriculture comparison
• Switching agriculture parameters to Naresh's new numbers
• Incorporating GDDs and KDDs in the yield estimation

Having agriculture allow double-cropping will wait for another PR.

[jrising]

Okay, this is now plausibly ready! Here are the big pieces:

• Within analysis/agcheck/, agricultural numerical diagnostics: -- simulate.jl: Produces byyear.csv to compare yields by crop, and irrigation.csv to compare irrigation. -- compare.R: Reads the files produced by simulate.jl and makes some graphs -- tarandi.jl: Compares Tara's precipitation with crop-specific deficits. Currently just does their correlation (it's -.27, which isn't so bad).
• Within prepare/agriculture/, new data generation: -- degreedays.jl: Writes crop-specific GDD and KDD data files for 1949 - 2009. -- extractbayes.R: Writes crop-specific coefficient files from Naresh's Bayesian simulations
• Changes to Agriculture.jl: -- Reads in the Bayesian coefficients when they're available -- Applies the GDD and KDD data in counties where available (and has 0's elsewhere!)