KPDB - migrate build SQL to dbt

Unmet success criteria: 3, 4, 6 — 5 is partially met.

Current state (after #2580)

12 of the 30 .sql files in sql/ are dbt models. The other 18 are build SQL that bash/02_build.sh runs through psql.

Two macros back those models: longform_by_boundary and match_geometry.

Reference CSVs (corrections_main, corrections_project, corrections_dob, zap_record_ids) are still loaded by create_corrections.sql, not seeds — criterion 2 is unmet as well.

Gotcha for whoever picks this up

dbt_project.yml sets model-paths: ["sql"], so every .sql file under sql/ is parsed as a model. dbt reports 30 models even though 18 are raw DDL it never runs. They resolve through ref() only because psql happens to have created tables of the same name first. A half-migrated file sitting in sql/ is silently in the graph, so partial migration is easy to get wrong here.

Start with dbt_project.yml, the 30 files under sql/, bash/02_build.sh, and create_corrections.sql. Map the 18 remaining build SQL files and the correction CSV loading before changing the dbt graph. Done means the remaining build work and correction data satisfy the migration criteria without half-migrated files being parsed as models.