feat: v202608 release

[jeffseif]

Background

This rolls up a number of changes for the v202608 release

Changes

1. The concept of continents has been removed. Although they helped to manually track backfills, they bundled a lot of risk into individual runs and could result in hours of compute being thrown away. Now, the harmonize, emit, and attribute batches all operate (and cache) on individual tiles. This also means that when a country falls entirely within one tile, its results can be generated end-to-end in minutes -- rather than spending hours pre-generating results for the continent it resides in.
2. The list of 220 countries is now generated programmatically and materialized in a json.
3. The JD approach was attributing emissions to 11 crops, but only 3 of them were backed by production -- so eight were lacking emissions factors. NASS quickstats was expanded to include those eight additional crop productions and now JD emissions factors are generated for all 11 crops in the USA
4. FAOSTAT production is ingested in order to support validation scripts. The GFW TCL dataset is also ingested.
5. Tooling to measure drift across commits was added in tools/measure-drift.py
6. Validation tooling (for comparing with independent datasets) was added in validation. This is quite complex but now enables a transparent and repeatable process that can help to identify (and sometimes explain) critical differences with other datasets.
7. Several fixes and improvements were applied to the SLUC model. These mainly are caused by varying granularity of the crop taxonomy over the MAPSPAM snapshots.