Task: Seed the 2025-2027 FOMC meeting dates from federalreserve.gov

Seed a real 2025-2027 window of FOMC meeting dates as calendar_event rows (diary_entry_type central_bank_meeting) on the US.FOMC calendar, transcribed from the Fed's published meeting calendar. No formal feed exists – transcription with the citable source, per the Calendar Events doc survey.

• The 2025, 2026 and 2027 FOMC meeting dates are entered as calendar_event rows on US.FOMC with diary_entry_type central_bank_meeting.
• Each meeting is one calendar_event row dated the second day of the Fed's published two-day range — the statement/decision day, usually a Wednesday (settled 2026-08-10).
• The 2027 schedule is fully published (Fed release 2025-09-05, including the first 2028 meeting); all three years seed 8 meetings each.
• Every row carries source='federalreserve.gov'.
• The per-year count matches the Fed's published schedule (8 regularly scheduled meetings per year); each transcription is cross-checked against federalreserve.gov/monetarypolicy/fomccalendars.htm.
• The seeded dates drive the FOMC_MEETING schedule lookup from the unified-tenor-resolution task (2F resolves to the second meeting on-or-after spot, etc.).

(Implementation strategy. Written when work starts; key decisions are distilled into the parent story's * Decisions at close, but the plan itself stays — it is the historical record of what we did.)

• The August 2025 one-day notation vote is not a two-day meeting and is deliberately not seeded (it has no statement day on the two-day convention).
• RLS gap found while validating: calendar_events, diary_entry_types (D1) and tenor_schedules (D2) lacked row-level-security blocks in refdata_rls_policies_create.sql (hand-maintained file; codegen does not emit RLS). Three blocks added; validate_schemas.sh now reports zero warnings. The fix rides in this commit because the validator is the gate for the populate script's table.
• DB-based verification (apply create + populate to a scratch database) was not possible in this environment: the postgres superuser password is not in .env and peer auth fails, so the recreate-from-scratch flow cannot run. Verification is static: dates transcribed from the Fed's published calendar page (fetched 2026-08-10), the 2026 second days match the committed D2 test dates exactly, every second day is a Wednesday, and validate_schemas.sh passes.

Manual QA scenarios (scaffolded via compass add test_scenario, run through the QA Validation Runner panel) that verify this task. Link new ones here as they're created; the scenario doc itself links back via its "Verifies task" field.

Done. refdata_calendar_events_populate.sql seeds the 2025-2027 window: 24 rows on US.FOMC (diary_entry_type central_bank_meeting), one per meeting dated the second day of the Fed's published two-day range (statement day, a Wednesday in every case), each carrying source='federalreserve.gov' and a description quoting the full published range. The insert is idempotent on the natural key (tenant, calendar, date, entry type) and the script's summary section reports per-year counts (8 each). The script is wired into foundation_populate.sql after refdata_calendars_populate.sql, matching D1's placement of US.FOMC itself.

Dates transcribed from the Fed's published meeting calendar: 2025 (29 Jan, 19 Mar, 7 May, 18 Jun, 30 Jul, 17 Sep, 29 Oct, 10 Dec), 2026 (28 Jan, 18 Mar, 29 Apr, 17 Jun, 29 Jul, 16 Sep, 28 Oct, 9 Dec – the same dates the D2 tests resolve against), 2027 (27 Jan, 17 Mar, 28 Apr, 9 Jun, 28 Jul, 15 Sep, 27 Oct, 8 Dec).