[jeffseif]

Background

This is mainly a refactor to get the S and JD attributions lined up to a common approach before we implement trace for JDP. This was originally going to be one big 'ol branch but this preparation work became larger.

The changes

1. 23ca2eb Move S livestock production from attribute to trace. The design was to have attribute only allocate emissions and for trace to join those onto productions and rollup over jurisdictions. When we implemented livestock, it was convenient to do all of livestock's production within attribute since that depended on both map data (GPW's head raster) and tabular data (FAOSTAT's livestock patterns). We wanted to avoid passing head counts over time as new columns in attribute, causing JD and S attribute schemas to drift apart ... but today we decided to swap that tradeoff.
2. b1bdf22 Stand up JD's mapping from USDA/NASS/CDL crops to FAOSTAT commodities, including three negative custom ones for DROPPED, UNATTRIBUTED, and PASTURE which understandably don't have items in FAOSTAT. So far this isn't used
3. 3fea6ad ^^ for S (including custom negative ids for OTHER_OILCROPS and OTHER_PULSES) and then plug the new taxonomy mappings into trace. This also changes the schema for trace so that the pkey switches from commodity_name: str to commodity_id: int, causing the former to become a fact column with the FAOSTAT name. Also ensure the tools are brought along
4. f58ed2b have validation use the taxonomy mappings too
5. af2cb8f update docs to reflect all of this

Manual testing

1. Ran trace for HND and verified its schema and new FAOSTAT values
2. Compared numerical values against recent caches and found they matched within float epsilon

[blallen]

did a peer program, looks good