GTC-2826 Colombia-specific analysis in forest_change_diagnostic

[danscales]

GTC-2826 Colombia-specific analysis in forest_change_diagnostic

We now do country-specific analyses for Brazil (BRA), Argentina (ARG), and Colombia (COL). Updated the gfwpro_forest_change_regions dataset to new v20240529 version that includes a bit indicating a pixel location is in COL, so we can do the necessary Colombia-specific analysis.

Added a Colombia country-specific classification from the new Colombia col_frontera_agricola/v2024 dataset that I added. For Argentina, the analogous country-specific classification is OTBN.

For the GFWProCoverage layer, I changed the code to just return the actual sets of bits, rather than a map corresponding to all the 1 bits. It seems very inefficient to create a new Map object for every single pixel being scanned, whereas a bitmap with access functions is very efficient.

New columns:

• country_area - amount of area of location in each of BRA, ARG, and COL. Tells the front-end whether a country-specific tab needs to be shown for this location/list.
• classified_region_area - for COL/ARG only - area breakdown by COL (Frontera agricola) or ARG (OBTN) classifications. Eventually arg_obtn_area will be removed, since replaced by this more general column.
• tree_cover_loss_by_country_classified_region_yearly - for COL only, TCL by Frontier Agricola categories and by year

Existing columns that are extended to COL

• protected_areas_by_category_area - for ARG already and now COL
• landmark_by_category_area - for ARG already and now COL
• tree_cover_loss_by_country_yearly - for ARG and BRA already and now COL
• tree_cover_loss_by_country_wdpa_yearly - for ARG and BRA already and now COL
• tree_cover_loss_by_country_landmark_yearly - for ARG and BRA already and now COL

To support classified_region_area, I had to add a new two-level categorization type, ForestChangeDiagnosticDataDoubleTwoCategory (i.e. there are two levels of categorization by strings, and the final value is a Double). Added extra include checks in the fill operations of several of these data types, so we don't have an empty country entry when there is no forest loss, protected area, landmark, etc for a particular country.

Added a new Colombia-specific test. (Removed "*.tsv" from .gitignore - makes it confusing when you want to add new tsv input files.)

[danscales]

No comments really, looks good to me! Although curious if you've noticed any performance implication as we've added all these regional datasets.

Because of the laziness that I put in for FCD and also the fact that the datasets are regional, we only load tiles for the regional datasets when analyzing COL, ARG, or BRA. For example, we were already loading PRODES deforestation info only for Brazil. So, I don't expect any performance implications for locations that are not in COL, ARG, or BRA.

Thanks for the review!