Pad ARI ids in the equivalencies export to match the ontology

[KrishnaTO]

Stacked on #61 — its base is claude/remove-icd9-codes-from-data, so review that one first. GitHub will retarget this to main automatically once #61 merges. Stacking avoids a delete-versus-modify conflict on the four rows both changes touch.

What and why

69 rows in mappings/ari.equivalencies.tsv spelled their disease as ARI 1001 where the ontology and mappings/ari.sssom.tsv both spell it ARI 0001001. The three files therefore did not join on disease id without normalising first — which is exactly what kept the drift between the two exports invisible for so long.

12 distinct diseases, 69 rows, one column.

The correct spelling is read, not assumed

The fix takes each id from ARI_ID in ontologies/ari_t1d.owl rather than zero-padding to a fixed width. That matters: ARI:0002 and ARI:0003 are genuinely four digits in the ontology, and padding them to seven would have introduced the same class of bug in the other direction.

Verification

• 69 insertions, 69 deletions. On every rewritten row, column 2 is the only column that differs — checked mechanically, not by eye.
• No row added or removed.
• Whole-repository errors 94 → 25 on top of Remove 89 ICD-9 codes filed under ICD-10 fields #61; every subject-padding finding cleared.

Advances #23 ("Validate all diseases with database xrefs"); referenced rather than closed.

🤖 Generated with Claude Code