Task: Foundation-layer population for tenor reference data (tenor/tenor_anchor/tenor_convention)

Ship idempotent populate scripts seeding ores_marketdata_tenors_tbl, ores_marketdata_tenor_anchors_tbl, ores_marketdata_tenor_conventions_tbl, and ores_marketdata_tenor_convention_resolutions_tbl with the standard tenor set (O/N through 30Y, SPOT, and the special short-end labels) and the spot/forward and FX swap conventions (credit/CDS seeded as a convention row only, its resolution deferred — see the capture in Notes), plus the runtime resolver (resolve_end_date=/=resolve_window=/=windows_overlap) that reads these persisted rows instead of hardcoding tenor knowledge in code.

• Standard tenor rows (per tenor.org's table) and SPOT/TODAY/TOMORROW/NEAR_LEG/IMM_ROLL anchors are seeded
• Tenor convention rows (spot/forward, FX swap near-leg, credit/CDS IMM) and the documented tenor_convention_resolution membership/override rows are seeded
• Seeded via the Foundation layer (universal, always-present reference data — see Notes for why this is not a DQ Librarian dataset), not a new standalone bundle
• Populate scripts are idempotent and validate_schemas.sh passes
• A runtime resolver reads the persisted tenor/tenor_convention/tenor_convention_resolution rows to resolve a tenor to a date — no per-label knowledge hardcoded in code

(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.)

Originally scoped as a DQ Librarian dataset (catalog + dataset + artefact table + bundle membership, per the synthetic.fx_spot_configs=/ =refdata.currency_pairs pattern). Investigation of refdata_day_count_fraction_types_populate.sql showed that pattern is for optional, party-selectable seed data a user chooses to Apply — not what standard tenor labels are. Tenors are universal, always-needed reference data, same category as day-count conventions, so this task instead seeds directly into the Foundation layer (foundation_populate.sql), which every new environment gets unconditionally with no separate Librarian step.

This task also closes the loop flagged while reviewing the salvaged tenor_period value type (from PR #1522, briefly raised as PR #1535 and then closed/folded in here): a standalone runtime type that hardcodes tenor-resolution knowledge (e.g. "O/N = 1 day") duplicates, and can silently diverge from, what the persisted model is meant to be the single source of truth for. The fix was schema, not code: adding tenor_convention.resolution_algorithm and tenor_convention_resolution.offset_unit=/=offset_multiplier so a SPECIAL tenor's convention-specific duration is data, not a hardcoded switch statement. The runtime resolver (ores.marketdata.api::domain::tenor_resolution.hpp) is now a pure function over those persisted rows, proven by a test that resolves the same tenor (O/N) to two different dates under two different conventions purely because the input rows differ.

tenor_convention.resolution_algorithm also anticipates credit/CDS's genuinely different IMM-roll-based resolution (IMM_ROLL, vs. the ANCHOR_OFFSET algorithm implemented here) without needing a schema change later — see the capture Implement IMM_ROLL tenor resolution for credit/CDS conventions for the deferred resolver implementation and CDS-specific seed data.

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.

PR #1539 review round (3 automated passes, issue-level comments, no line comments):

Shipped: four Foundation-layer populate scripts (tenor anchors, tenors, tenor conventions, tenor convention resolutions) seeding 5 anchors, 37 tenors, 3 conventions, and 39 resolution rows, verified idempotent against a freshly recreated dev DB (re-running produces identical row counts). Added the SQL create/drop/RLS-policy wiring codegen doesn't generate for a junction entity. Added the runtime resolver (resolve_end_date=/=resolve_window=/=windows_overlap) as a pure function over the persisted rows — 9 test cases / 11 assertions, including a test proving the same tenor resolves differently under different conventions. Credit/CDS is modeled and its convention row seeded, but its IMM_ROLL resolution is deferred to a capture, not implemented or fabricated here.