Rename the CLI's WorkflowResults to RunSummary

[isayev]

The collision

name	module	what it is
WorkflowResult	Auto3D/results.py	main()'s return value — a str subclass that is the output SDF path
WorkflowResults	Auto3D/cli/results.py	the CLI's display payload — counts, elapsed time, failure list

One trailing s apart, in two modules both named results.py, and not related types — nothing converts between them. Importing either into a file that held the other read as a typo in both directions.

RunSummary names what it is rather than where it came from. Auto3D.results.WorkflowResult — the one the Python API returns — is untouched.

The guard is a rule, not a forbidden name

test_no_class_name_differs_from_another_only_by_a_trailing_s rejects any such pair anywhere in src/, because the defect is the near-collision rather than this particular instance. Same treatment test_only_one_class_in_the_package_is_named_tautomer_engine already gives the TautomerEngine collision sitting beside it.

Mutation-tested by introducing an unrelated pair (FailedMolecule / FailedMolecules), which it catches and names — so it is checking the property, not remembering an answer.

One mechanical detail worth flagging

The rename used a word-boundary pattern, which matters more than usual here: WorkflowResults contains WorkflowResult as a prefix, so an unanchored replace of the shorter name would have silently corrupted every site of the longer one. Only the longer name changed, and there are zero residual references to the old name in src/ or tests/.

Compatibility

Still an importable public name, so it gets a CHANGELOG table like the isomer_engine split, and no shim. It appears nowhere under docs/source, so no documentation changes.

Verification

• 1770 passed, 1 skipped, 74 deselected (+1 new guard), randomized order.
• mypy: 68 errors in 22 files, 75 checked — unchanged.

Context

First piece of wave 9. The remaining two are the layer-layout regroup and splitting ASE/thermo.py (1895 lines — now the largest file in the tree).