feat(swarm): AgentClient topology apply/teardown/state methods

Three new RPCs mirroring the existing deploy/teardown/status pattern:

- apply_topology(hydrated, version_hash) — long-timeout (600s) for
  image pulls + compose up.
- teardown_topology(topology_id) — 300s timeout; enough for a
  stubborn compose-down without hanging a heartbeat.
- get_topology_state() — short control-plane read for reconcile.

The per-call timeout swap uses the same trick as .deploy().

Step 5 of the agent <-> topology integration.

decnet/swarm/client.py

@@ -34,6 +34,11 @@ log = get_logger("swarm.client")
 # later iteration if the default proves too short.
 _TIMEOUT_DEPLOY = httpx.Timeout(connect=10.0, read=600.0, write=30.0, pool=5.0)
 _TIMEOUT_CONTROL = httpx.Timeout(connect=5.0, read=15.0, write=5.0, pool=5.0)
+# Topology apply pulls images + runs compose on the agent — same ball-park
+# as a fleet deploy.  Teardown is faster but still long enough we can't
+# reuse the control timeout.
+_TIMEOUT_TOPOLOGY_APPLY = httpx.Timeout(connect=10.0, read=600.0, write=30.0, pool=5.0)
+_TIMEOUT_TOPOLOGY_TEARDOWN = httpx.Timeout(connect=10.0, read=300.0, write=30.0, pool=5.0)
 
 
 @dataclass(frozen=True)
@@ -191,6 +196,51 @@ class AgentClient:
         resp.raise_for_status()
         return resp.json()
 
+    # ------------------------------------------------------------ topology
+
+    async def apply_topology(
+        self,
+        hydrated: dict[str, Any],
+        version_hash: str,
+    ) -> dict[str, Any]:
+        """Push a hydrated topology to the agent for local materialisation.
+
+        The agent independently computes ``canonical_hash(hydrated)`` and
+        returns 400 if it disagrees with *version_hash* — that's how we
+        catch serialisation drift before half-creating bridges.
+        """
+        old = self._require_client().timeout
+        self._require_client().timeout = _TIMEOUT_TOPOLOGY_APPLY
+        try:
+            resp = await self._require_client().post(
+                "/topology/apply",
+                json={"hydrated": hydrated, "version_hash": version_hash},
+            )
+        finally:
+            self._require_client().timeout = old
+        resp.raise_for_status()
+        return resp.json()
+
+    async def teardown_topology(self, topology_id: str) -> dict[str, Any]:
+        """Ask the agent to dismantle the named topology."""
+        old = self._require_client().timeout
+        self._require_client().timeout = _TIMEOUT_TOPOLOGY_TEARDOWN
+        try:
+            resp = await self._require_client().post(
+                "/topology/teardown",
+                json={"topology_id": topology_id},
+            )
+        finally:
+            self._require_client().timeout = old
+        resp.raise_for_status()
+        return resp.json()
+
+    async def get_topology_state(self) -> dict[str, Any]:
+        """Snapshot of the agent's applied topology + live docker state."""
+        resp = await self._require_client().get("/topology/state")
+        resp.raise_for_status()
+        return resp.json()
+
     # -------------------------------------------------------------- diagnostics
 
     def __repr__(self) -> str: