feat(agent): real worker-side /mutate with master swarm dispatch

- Implement /mutate handler: load_state, update services + last_mutated,
  save_state, write_compose, compose up -d via asyncio.to_thread. 404
  for missing state / unknown decky_id. dry_run short-circuits before
  any side effect.
- Add AgentClient.mutate(decky_id, services, *, dry_run=False) using
  _TIMEOUT_DEPLOY (compose up can pull/build, exceeds control timeout).
- mutator/engine.py: in swarm mode with decky.host_uuid set, resolve
  worker via _resolve_swarm_host and dispatch through AgentClient.mutate
  instead of writing a compose file on master. Master-resident deckies
  (unihost mode, or swarm with host_uuid=None) keep the local path.

decnet/swarm/client.py

@@ -256,6 +256,29 @@ class AgentClient:
         resp.raise_for_status()
         return resp.json()
 
+    async def mutate(
+        self,
+        decky_id: str,
+        services: list[str],
+        *,
+        dry_run: bool = False,
+    ) -> dict[str, Any]:
+        body = {
+            "decky_id": decky_id,
+            "services": list(services),
+            "dry_run": dry_run,
+        }
+        # Worker /mutate runs `compose up -d` which can pull/build; same
+        # long-tail latency as /deploy. Swap the deploy timeout in.
+        old = self._require_client().timeout
+        self._require_client().timeout = _TIMEOUT_DEPLOY
+        try:
+            resp = await self._require_client().post("/mutate", json=body)
+        finally:
+            self._require_client().timeout = old
+        resp.raise_for_status()
+        return resp.json()
+
     async def teardown(self, decky_id: Optional[str] = None) -> dict[str, Any]:
         resp = await self._require_client().post(
             "/teardown", json={"decky_id": decky_id}