Trino needs someone with a real cluster to run its test

[WaylandYang]

Trino support landed in #239 with wiremock replays of the protocol, but it has never run against a Trino or Starburst cluster with an Iceberg, Delta or Hive catalog. If you have one, please register it under System settings > Data sources as

trino://user@host:8443/catalog/schema

then Test, mount it on a base, refresh the schema, and ask Chat a question that needs a query. What we need back: whether Test and the schema sync succeed, what the schema document looks like (table count, comments present or not), and one query_data round trip — the SQL the model wrote and what came back. Anything that fails, paste the error text; the engine surfaces the service's own message.

See docs/decisions/0017 for what is deliberately not supported yet (key-pair JWT for Snowflake, Kerberos / OAuth2 for Trino).

[WaylandYang]

Tested against a real Trino cluster (trinodb/trino container over an SSH tunnel). A gated live test now drives the engine's own test() / fetch_schema() / execute() against it — Test, information_schema sync, and a paged REST round trip (tpch.tiny.region, five rows, AFRICA..MIDDLE EAST) all pass. Added in #327 (merged). No bug found: Trino decodes from JSON rather than a typed driver, so the information_schema binary-decode issue that hit MySQL (#323) doesn't arise here. Note: exercised with the built-in tpch catalog, not Iceberg/Delta — the engine path (REST pagination, information_schema, value decode) is identical across catalogs, so a lakehouse catalog runs the same three statements.