ontologies: no stray-header guard on the nodes path (symmetry gap)

Found in the adversarial review of #704. LOW — a symmetry gap, not a live defect. #704 added a backstop dropping edge rows whose `subject` cell is literally `"subject"`. `_normalize_schema` also handles the **nodes** file, which has no equivalent guard. Verified currently harmless: all 14 `*_nodes.tsv` files have zero stray `id` rows, and the `ec`/`mondo`/`upa` nodes branches never prepend a canonical header the way the ec edges branch does — so there is no producer for the defect on that side. Worth adding for symmetry if a nodes branch ever gains a header-prepending step, and cheap to do (`df[ID_COLUMN].astype(str) == ID_COLUMN`). Filed so the asymmetry is a recorded decision rather than an oversight.