fix: correct return type annotation of get_nsys_entrypoint

[andrewwhitecdw]

Bug

get_nsys_entrypoint returns a tuple ("nsys", ""), but its annotation was str. This mismatched the implementation and callers that unpack the result.

Fix

Update the return type annotation to tuple[str, str].

Test

The existing TestExecutor::test_get_nsys_entrypoint already asserts the tuple return value.

Verification

uv run pytest test/core/execution/test_base.py::TestExecutor::test_get_nsys_entrypoint -v passes. uv run --group lint ruff check ... and ruff format --check ... pass.

[svcnemo-autobot]

PR #568 — needs attention
Blocking finding: nemo_run/core/execution/slurm.py:564-570 still annotates its override as returning str, although it returns the same two-string tuple as the corrected base method and is tuple-unpacked by nemo_run/run/torchx_backend/packaging.py:253. Update the override to tuple[str, str] so concrete SlurmExecutor use is typed correctly.
Non-blocking findings: None.
Validation gaps: Runtime tuple tests exist, but they do not catch this annotation mismatch. Only DCO is reported.

[ko3n1g]

hey @andrewwhitecdw, could you fix the type hint? I'll get it merged afterwards. Thanks for the contribution!

[andrewwhitecdw]

Fixed the same mismatch in the Slurm override as well. Both the base Executor and SlurmExecutor now annotate get_nsys_entrypoint as tuple[str, str], matching their implementations. CI lint/test passed locally.

[ko3n1g]

/ok to test

[ko3n1g]

Thank you!