Prism Refresh Readiness

Freshness gate for the 15-minute recompute tick: which prisms have new data worth paying for? Batch — one call covers every live prism.

Per prism: resolve the granted tables at compute-time (the ontology edge is the authority; the set is a function of the whole slice and shifts on ontology rebuilds), read each table's table_max_date data edge, and compare it against that table's own last-consumed edge. Strict ALL: ready only when every granted table advanced — a series must never gain a day computed from a subset of its sources while history used all of them. Neo4j + PG only — no warehouse touch. Fail-open per table on any missing signal (new table, no edge row), and per prism when nothing resolves.