-- 0.2.0 -> 0.2.1
--
-- Nothing to change in the schema: both fixes live in the library (verify.rs), and they
-- reach a database the moment the new .so is loaded. The script exists so the catalog
-- says which version is running -- an extension whose behaviour changed under the same
-- version number is the failure this family of extensions exists to close.
--
-- What 0.2.1 refuses at propose, both found by a cycle harness with a superuser watching
-- (an LLM proposing through the gate against a two-tenant database, 2026-10-06):
--
-- 1. A CTE THAT WRITES. max_rows counted the rows of the outer statement only, so
--    `with d as (delete from t returning id) select count(*) from d` deleted every row
--    under a limit of 5. Each write can still be proposed as its own statement.
-- 2. set_config() ANYWHERE IN THE STATEMENT. The session allowlist judged SET and the
--    startup parameters, but while the gate runs a proposal its hooks step aside, so
--    `... where set_config('app.tenant_id', '2', true) is not null` read another
--    tenant's rows. Matched by OID in the analyzed and rewritten tree: where, from, a
--    sublink, a CTE, a view, schema-qualified or not.
--
-- Still NOT covered, and said in the README: a function that already exists and calls
-- set_config in its own body.