Get Organization Memory

Customer memory for one organization, given directly (org_id) or by a ticket that belongs to it (ticket_id) -- Frederick: "he should be able to look at a customer's history... has this customer had this complaint in the past." This is what makes that history reachable from a Rooster surface that has no ticket in hand at all (issue intake, the CS directive, playbook authoring, the direction line), not only the CS-draft recipe that already reads it inside get_cs_ticket_context above.

Gated on DIRECTION_WRITE, the same "this caller is a Rooster agent, not a human UI" grant get_clock and the issue-intake/playbook-authoring reads already reuse below -- this route's own scoping (disclosure.for_customer/ graph_registry.resolve_customer_graphs, inside agent_draft.build_organization_memory) is what keeps every response customer-safe and sealed to exactly one org, never which grant a caller happens to hold.

Exactly one of org_id/ticket_id is expected; if ticket_id is supplied it resolves to THAT ticket's own org, and query defaults to that ticket's own subject + latest customer message when left blank. If neither resolves to a real organization -- no id given, an unknown ticket, a ticket with no org (B2C), or an unknown org -- this returns an empty bundle rather than a 404: "if a tool cannot resolve an org, it returns nothing rather than searching broadly," never an error an agent might retry into a wider search.