Fix SOFA CV scoring, urine_output_rate diffs, and Elixhauser no_drg

[Chessing234]

Summary

• SOFA CV score 3 now requires epi/norepi rate > 0 (rate 0 no longer scores 3)
• urine_output_rate uses second-based DATETIME_DIFF so BigQuery and Postgres agree (DATETIME_DIFF implementation varies between postgres and bigquery #1549)
• Elixhauser AHRQ no_drg excludes primary diagnoses as documented and RTRIMs ICD-9 codes (Spaces in codes in elixhauser_ahrq_v37_no_drg.sql #1077)

Test plan

• pytest tests/ for changed concepts (sqlfluff + transpile)
• Re-transpiled postgres/duckdb dialects for edited BigQuery sources
• CI transpile / concepts workflows

Made with Cursor

[Chessing234]

Fixes SOFA CV zero-rate scoring, urine_output_rate BigQuery/Postgres drift (#1549), and Elixhauser no_drg primary-dx + ICD-9 trim (#1077).

[Chessing234]

Concept changes from the old mega-PR are landing here / in follow-ups; load-path/transpile/CI moved to #2127.

[Chessing234]

SQLFluff itself is fine — the red check was the annotate action lacking fork permissions. Pulled in the same ignore-unauthorized-error fix as #2128.

[Chessing234]

dropped the continue-on-error sqlfluff hunk — that's in #2148 now. this one is concept-only again (sofa cv, uo rate, elixhauser). @alistairewj ready for review when you have a minute.