Note

Medium Risk
Touches client-side cursor iteration and connection-provider behavior; correctness depends on proper node selection and cursor lifecycle handling in both standalone and cluster deployments.

Overview
Introduces AggregateIterator, a new cursor-based iterator for FT.AGGREGATE that keeps cursor reads/deletes on the same Redis node and borrows/returns pooled connections per batch to avoid pool exhaustion/leaks; UnifiedJedis and RediSearchCommands gain ftAggregateIterator(...) to expose it.

Deprecates legacy iteration/cursor APIs (FtAggregateIteration, FtSearchIteration and UnifiedJedis.ftSearchIteration(...), plus ftCursorRead/ftCursorDel in RediSearchCommands) and extends AggregationBuilder with cursor batch size tracking plus AggregationResult.isEmpty() to support iterator semantics; adds standalone+cluster integration tests and unit tests, and fixes PooledConnectionProvider.getPrimaryNodesConnectionMap() to return the pool rather than borrowing a connection.

^{Written by Cursor Bugbot for commit 464d231. This will update automatically on new commits. Configure here.}