fix(executor): partition grouped global batches

[sylvesterkaczmarek]

Summary

• keep grouped global-batch operators hash-partitioned even when actor concurrency is 1
• retain at least the dataset's existing block count so unrelated source groups are not coalesced into one oversized Arrow block
• preserve single-block behavior for truly ungrouped global-batch operators
• add regression coverage for existing block counts, actor-pool concurrency, and ungrouped stages

Fixes #2643

Validation

• python -m compileall -q nemo_retriever/src/nemo_retriever/graph/executor.py nemo_retriever/tests/test_global_batch_partitioning.py
• git diff --check

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[greptile-apps]

Confidence Score: 5/5

[Medium risk] Refactors data partitioning logic in the execution engine.

The PR appears safe to merge based on the changes in its head-versus-base diff.

Summary

The PR hash-partitions grouped global-batch stages while retaining at least the dataset’s existing block count. Ungrouped stages still use one block. New tests cover existing block counts, actor concurrency, and ungrouped behavior.

_{Reviews (4) · Last reviewed commit: "fix(executor): partition grouped global ..."}

[sylvesterkaczmarek]

Refreshed this PR onto current upstream today. It is now 0 commits behind and mergeable, with the intended patch preserved and no unresolved review threads. This has been quiet for over two weeks. Could a maintainer please review/merge it, or let me know if anything else is needed?

[sylvesterkaczmarek]

Rebased onto current main today; the new GitHub-Verified head is c83151f, 0 behind and mergeable, with no failing checks or unresolved review threads. @nkmcalli, could you review the refreshed head when convenient?