feat(query): consolidate dynamic hydration into single UNION ALL query

Michael Angelo Riverarequested to merge
michaelangeloio/clickhouse-optimizations-hydration-1 into main

Replaces N separate compile+execute hydration queries (one per discovered entity type) with a single consolidated UNION ALL query. Each arm does a primary-key point lookup with toJSONString(map(...)) for uniform column alignment across heterogeneous entity types.

Before: 1 base + N hydration queries (N = number of distinct entity types in results, typically 3-7) After: 1 base + 1 consolidated hydration query

Profiled against dev ClickHouse Cloud (27M edges, 1.5M users):

  • Neighbors (6 entity types): 7 queries → 2 queries, hydration 700ms → 123ms
  • PathFinding (3 entity types): 4 queries → 2 queries, hydration 259ms → 91ms
  • 13 entity types in single UNION ALL: 68ms, 19K rows read, 25MB memory

Merge request reports