fix(query-orchestrator): align pre-aggregation refresh key cache TTL with renewalThreshold

[igorferst]

Summary

Pre-aggregation refresh key queries were cached with a hardcoded 1-hour Redis TTL in PreAggregationLoadCache.keyQueryResult, while refresh_key.every only controlled renewalThreshold. When every was longer than one hour (e.g. 15 days), the cache entry expired hourly, so refresh-key SQL still ran about every hour despite the model config.

This PR sets cache expiration to max(renewalThreshold, 1 hour) using the same resolved threshold as today (refresh_key.every, or queryCacheOptions.refreshKeyRenewalThreshold).

Why

• renewalThreshold and Redis TTL were out of sync: stale logic could allow 15-day reuse in theory, but the key was removed after 1 hour → cache miss → refresh-key SQL re-executed.
• Math.max(renewalThreshold, 60 * 60) fixes long every values and preserves the previous minimum 1-hour TTL for short thresholds (sql without every, default ~10s renewal, etc.).

Behavior changes

Config	Before	After
sql + every: '15 days'	Refresh-key SQL ~hourly	Cached ~15 days
sql only (renewalThreshold ~10s)	1h Redis TTL	Still ≥1h Redis TTL
refreshKeyRenewalThreshold in orchestrator	Overrides renewal only	Overrides both renewal and expiration

Made with Cursor