run_aggregate.cpp is 1,652 lines and mixes QO-owned pipeline orchestration with QE-owned cursor and batch execution machinery. Extract the QE-owned helpers into a new run_aggregate_exec.cpp / run_aggregate_exec.h pair, and assign the new files to 10gen/query-execution in OWNERS.yml.

Functions to extract: registerCursor, collectQueryStats, getFirstBatch, executeSingleExecUntilFirstBatch, executeUntilFirstBatch (and optionally handleMultipleCursorsForExchange).

The remaining pipeline orchestration, view resolution, and explain logic stays in run_aggregate.cpp under 10gen/query-optimization.

As a stretch goal (or follow-up), identify QI-owned functions still remaining in run_aggregate.cpp and either move them into a run_aggregate_qi.cpp in the same PR or file a follow-up.

Likely candidates include computeShapeAndRegisterQueryStats (query stats/observability, QI), and createMongotMetadataPipelineIfNeeded / getDocResultsAndMetadataStage / addMetadataCursorExecutor (search extensions, QI-search).