The schema was added before the Hindi (73b0c68) and Polish/Korean
(118e4bd) instructions landed, so languageMap and languageStepsMap
still only declared en, es, it, tr, ru, zh.
All 1,324 records carry all nine languages and the README documents
nine, but `additionalProperties` let hi/pl/ko through unvalidated and
unrequired — a record silently missing them still passed validation.
Declare all nine in `properties` and `required` for both maps.
Verified: the 1,324 records validate cleanly against the updated
schema, and a record missing hi/pl/ko is now rejected.
Adds data/exercises.schema.json (Draft 2020-12) describing every field,
its type and constraints, so the dataset and community additions can be
validated with any standard JSON Schema validator. Validated against all
1,324 records with zero errors.
Also documents the previously undocumented instruction_steps field and
references the schema from the README.
umutcakirai