Give Instagram connector deletes the long cascade-delete timeout - #1641

Summary

Deleting an Instagram connector runs two statements in one Prisma transaction: the tool-call log delete (added by #1630 for the Meta Platform Terms) and the connector delete, which cascades into tools, graph nodes and edges. On 11-11 at 16:17 UTC one such delete took 5,977 ms, Prisma rolled it back at its default 5 s timeout and the user got a 500 (Sentry ANYTHINGMCP-CLOUD-BACKEND-4).

#1639 removed the main cost (kg_edges had no index on its node columns). A connector with many log rows can still pass 5 s on the log delete alone, so this transaction now uses CASCADE_DELETE_TRANSACTION (60 s), as account and workspace deletes already do.

Changes

• ConnectorsService.remove(): pass CASCADE_DELETE_TRANSACTION to the Instagram $transaction([...]).
• Spec: the transaction is called with those options and their timeout is above 5 s.

Type

• Bug fix

Testing

• Added/updated unit tests
• npx jest src/connectors/connectors.service.remove.spec.ts in packages/backend: 3/3 passed. Full lint, typecheck and tests run in CI.

Cloud and self-hosted behave the same: only the timeout of this one delete changes. No schema change, nothing in ee/.

Checklist

• My code follows the existing code style
• I have tested my changes locally
• All existing tests pass (npm test) (CI)
• This PR has a descriptive title

Related Issues

Relates to #1638 and #1630.