Expose script's actors client to trusted spawned actors

[NotASithLord]

Problem

The JavaScript-first orchestration path is currently inconsistent:

• a trusted-lineage spawned actor granted message_actor may delegate directly and receives the reply in-band;
• the same actor granted both script and message_actor gets no actors.* client inside script, because the tool mint and the actors/call route reject every kind: 'spawned' owner.

That forces the model back into one-tool-call-per-turn choreography exactly where code should be strongest: fan-out, chaining replies, retry/timeout logic, and returning only the synthesized result.

Direction

Expose the existing actors.list/ask/send client to a spawned actor only when its already-narrowed grants contain the corresponding authority. Reuse the current messageActor sender gate on every op; do not add a second delegation path.

This is an ergonomics change, not an authority expansion:

• require both script and message_actor for ask/send;
• require actor_list for list;
• preserve the trusted-lineage and inbound-turn checks;
• keep bound environment actors blocked from cross-actor delegation;
• preserve root-keyed rate/outstanding caps, dedupe, audit, fencing, tracing, and Stop cancellation.

Acceptance criteria

• Trusted spawned actors can fan out/choreograph actor work inside one script call.
• A spawned actor without message_actor cannot mint or use the client.
• Tainted/untrusted lineage is refused by the existing sender gate.
• actors.list() cannot bypass the spawned actor's narrowed grant set.
• Bound actors remain unable to delegate to other actors.
• Focused unit coverage proves minting, per-op authorization, lineage refusal, and Stop behavior.