fix(api/ontologies): offload blocking filesystem IO off the event loop

[ly-wang19]

What

The delete_ontology and list_ontologies ontology routes are async def, but call the synchronous OntologyService methods directly:

@router.delete("/{ontology_key}")
async def delete_ontology(...):
    ontology_service.delete_ontology(...)   # blocking stat/unlink on the event loop

@router.get("")
async def list_ontologies(...):
    ontology_service.list_ontologies(user)  # blocking metadata read on the event loop

Both service methods do blocking filesystem IO (Path.exists/resolve/unlink, metadata read), so they stall the event loop for the duration. The sibling upload_ontology route already awaits its async service method — these two were the inconsistent blocking paths.

Change

Dispatch the two synchronous calls via asyncio.to_thread. The service methods stay synchronous on purpose — their existing unit tests (tests/unit/api/test_delete_ontology.py) call them directly — so only the async routes change.

Test

Adds tests/unit/api/test_ontology_router_offload.py: extracts each route handler and asserts the blocking call is dispatched through asyncio.to_thread. Reverting either offload fails the corresponding test.

pytest tests/unit/api/test_ontology_router_offload.py tests/unit/api/test_delete_ontology.py → 10 passed; ruff clean.

[github-actions]

Hello @ly-wang19, thank you for submitting a PR! We will respond as soon as possible.

The base branch was changed.