Normalize flow outputs and retain dynamic subflow jobs

[Andrew-S-Rosen]

Summary

• Normalize nested Job and Flow values returned by @flow into output references.
• Collect all jobs created while a job function executes when its return value defines a dynamic subflow.
• Preserve the full dynamic subflow as the replacement flow instead of retaining only the returned terminal job.
• Add regression tests for nested decorated-flow outputs and multi-job dynamic subflows.

Why

Downstream workflow adapters currently need to wrap decorated flow outputs recursively and compose @job with @flow to prevent intermediate jobs in dynamic subflows from being lost. This behavior is general Jobflow composition logic and belongs upstream.

Together with #885, this allows Quantum-Accelerators/quacc#3405 to remove its Jobflow-specific normalization and subflow patches from src/quacc/wflow_tools/decorators.py.

Additional dependencies introduced

None.

Validation

• PYTHONPATH=src python -m pytest tests/core/test_flow_decorator.py tests/core/test_job.py -q — 39 passed, 1 xfailed
• python -m ruff check --ignore PLR0917 src/jobflow/core/flow.py src/jobflow/core/job.py tests/core/test_flow_decorator.py tests/core/test_job.py — passed
• git diff --check — passed

Checklist

• Tests have been added for the new behavior.
• Focused linting and tests pass.

[Andrew-S-Rosen]

Superseded by #885. The relevant changes and regression tests have been consolidated there using a single shared normalization helper.