Update cait datasets

[pabloarosado]

• Create a script to download and prepare CAIT datasets.
• Add a generic module for sanity checks.
• Add script to run sanity checks on CAIT datasets.
• Create a script to upload CO2 dataset files to S3.
• Update the main script to generate CO2 dataset files.
• Add CAIT dataset files to repos (in a new "grapher" folder).
• Update the CO2 dataset csv file and codebook, and remove the json and xlsx files (they will be hosted only in S3).
• Update README (to have links to the CO2 dataset files in S3).

[pabloarosado]

Thanks @bnjmacdonald, I was trying to follow the philosophy of the energy-data repos. But I agree with you that it would be better to use importers or etl to upload the CAIT datasets to grapher. I could either do it now and cancel this PR, or leave it for a future improvement. I don't have a strong opinion on which option is better, @edomt ?

[edomt]

Indeed the general pipeline isn't optimal – but of course, it was already like that before. There are several things to take into account, including that:

• importers will be deprecated in the coming months
• etl isn't fully ready yet, so even an implementation of this pipeline in etl would need to be revisited at some point
• there are a lot of outstanding & high-priority datasets to deal with.

What I would suggest to get rid of the confusion mentioned by @bnjmacdonald is that:

• we move prepare_cait_datasets.py & its output into a cait folder in importers in its current state, i.e. without refactoring it as a "true" importer that upserts data into Grapher (because that part will be deprecated soon). So something that looks like the population folder.
• The output of that script is then manually uploaded to Grapher
• main.py in this co2 repo then fetches that Grapher dataset to use it

And at the next update of CAIT (presumably next year?), we'll transform the whole thing into a proper etl pipeline.

Let me know what you think :)