add CPP housing growth outputs by NTA and community district

[damonmcc]

related to #2578
builds on this branch

New outputs

Two CSVs for the Capital Projects Portal, by NTA and community district, one row per geography.

Column	What it is
units_2020_census	2020 Census count, as of April 1 2020
units_2020	units at the end of 2020: the census count plus the rest of that year
completed_units_2016_2025	net units completed over the past 10 years
completed_units_2021_2025	net units completed between units_2020 and units_2025
units_2025	units_2020 plus the line above
projected_completed_units_2026_2035	KPDB units phased within the next 10 years
projected_units_2035	units_2025 plus the line above

Year-labeled counts are as of the end of that year, so a project completed in 2025 counts toward 2025. Both completed_units_* columns are net of demolitions and lost units rather than gross completions.

Data corrections

• Projects that match no boundary keep their unit measures. The fallback that assigns them to an intersecting geography only set the geography, proportion, and units_net, so about 1,300 projects per geography contributed nothing to any phasing total.
• Units are split by largest remainder instead of rounding each geography's share on its own, so a project's parts sum to its total. NTA and CD now report the same citywide numbers, where they used to differ.
• Hudson Square gets its corrected geometry (cherry-picked from KPDB SCA and MTA enhancements #2095). ZAP has P2012M0255 at 2.1 km² with its centroid about 6 km from where it belongs. The corrected shape is 0.2 km² and lands in MN0201 SoHo-Little Italy-Hudson Square.

Source data

• ZAP moves from 20250303 to 20260817. DCP Applications used to stop at 2025-02-18 while DOB ran to 2026-01-20, so nothing filed in the previous 18 months was in KPDB at all.
• Housing DB moves to 26Q2.
• assert_project_and_dob_data_align warns when the two drift apart again.

Infrastructure

• Output goes to edm-publishing through the standard publishing connector, landing in db-kpdb/build/<build> like every other product. Replaces a KPDB-specific upload.py that wrote datetime-stamped folders to edm-private.
• Build inputs are dbt sources instead of schema-qualified strings, and zap_project_many_bbls and summary_record_phasing are dbt models now.
• A README covering what KPDB is, how outputs get calculated, and what a record, cluster, and project are.

Notes

The phasing buckets are relative, not calendar dates: DOB phasing follows job status, so it moves every build, and shares that come from a source were written whenever that file was made.

units_2020_census is the only fixed number. Everything else moves with the dcp_housing version.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 68.57%. Comparing base (96e1b9f) to head (eb3ea2b).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.