Task: Implement Hotfix: the trading children write outside the session party

Give a model a way to say that its integration test writes under a party-scoped session when the codegen cannot derive that from the entity's own columns, and let the two trading children use it.

• The two eventing tests scope their session to a seeded party.
• ores.trading.core.tests passes.
• No other component's generated output changes.

Three changes:

1. core.py: party_scoped_writes also honours a model-declared :party_scoped_writes: true in the * SQL ** Flags drawer.
2. The eventing-test template: the v.party_id = *party_ctx.party_id() line is gated on has_party_column as well as party_scoped_writes, because a party-scoped session no longer implies the entity carries the column.
3. The two models set the flag.

The diagnosis took three steps, each proved against the running database:

1. The test aborted dereferencing an empty optional – party_ctx.party_id() – because party_scoped_writes had gone false for these two entities.
2. Seeding a party and writing the parent with it worked, and the notification chain proved out, so the writes themselves were sound.
3. The CRUD round trip then failed with already_exists. The cause is the replace claim: trade_state_repository::replace_claim reads the current row, the row-level security policy hid it (the session had no visible party), the read came back empty, and the claim therefore said "no row exists" for a row that does. The insert trigger, which gives version zero the one meaning "no current row exists", refused it.

So the missing piece was the session party, not the seed.

trade_state and trade_booking gained :party_scoped_writes: true; the two tests now run under a party-scoped session and ores.trading.core.tests is green.