perf: optimize gateway cold start from ~6.9s to ~460ms (#3918)

Channel lazy load: discover_enabled() only imports enabled channel
modules instead of all 18 modules with heavy SDKs (telegram, discord,
slack, etc). discover_all() now delegates to discover_enabled().

Lazy OpenAI client: defer AsyncOpenAI() + httpx construction to
_ensure_client() with asyncio.Lock double-checked locking. openai
and httpx imports moved from module-level into _ensure_client().

Minor: lazy Nanobot/RunResult and CronService exports via __getattr__.

Benchmark: 6910ms → 460ms (-93.3%)

tests/channels/test_channel_plugins.py

@@ -180,7 +180,7 @@ async def test_manager_loads_plugin_from_dict_config():
     )
 
     with patch(
-        "nanobot.channels.registry.discover_all",
+        "nanobot.channels.registry.discover_enabled",
         return_value={"fakeplugin": _FakePlugin},
     ):
         mgr = ChannelManager.__new__(ChannelManager)
@@ -210,7 +210,7 @@ async def test_manager_propagates_groq_transcription_api_base_to_channels():
     )
 
     with patch(
-        "nanobot.channels.registry.discover_all",
+        "nanobot.channels.registry.discover_enabled",
         return_value={"fakeplugin": _FakePlugin},
     ):
         mgr = ChannelManager.__new__(ChannelManager)
@@ -246,7 +246,7 @@ async def test_manager_propagates_openai_transcription_api_base_to_channels():
     )
 
     with patch(
-        "nanobot.channels.registry.discover_all",
+        "nanobot.channels.registry.discover_enabled",
         return_value={"fakeplugin": _FakePlugin},
     ):
         mgr = ChannelManager.__new__(ChannelManager)
@@ -498,10 +498,8 @@ async def test_manager_skips_disabled_plugin():
         providers=SimpleNamespace(groq=SimpleNamespace(api_key="")),
     )
 
-    with patch(
-        "nanobot.channels.registry.discover_all",
-        return_value={"fakeplugin": _FakePlugin},
-    ):
+    ep = _make_entry_point("fakeplugin", _FakePlugin)
+    with patch(_EP_TARGET, return_value=[ep]):
         mgr = ChannelManager.__new__(ChannelManager)
         mgr.config = fake_config
         mgr.bus = MessageBus()

tests/providers/test_local_endpoint_detection.py

@@ -85,17 +85,18 @@ class TestIsLocalEndpoint:
 class TestLocalKeepaliveConfig:
     """Verify that local endpoints get keepalive_expiry=0."""
 
-    def test_local_spec_disables_keepalive(self):
+    async def test_local_spec_disables_keepalive(self):
         spec = _make_spec(is_local=True)
         spec.env_key = ""
         spec.default_api_base = "http://localhost:11434/v1"
         provider = OpenAICompatProvider(
             api_key="test", api_base="http://localhost:11434/v1", spec=spec,
         )
+        await provider._ensure_client()
         pool = provider._client._client._transport._pool
         assert pool._keepalive_expiry == 0
 
-    def test_lan_ip_disables_keepalive(self):
+    async def test_lan_ip_disables_keepalive(self):
         """A generic 'openai' spec with a LAN IP should still disable keepalive."""
         spec = _make_spec(is_local=False)
         spec.env_key = ""
@@ -103,16 +104,18 @@ class TestLocalKeepaliveConfig:
         provider = OpenAICompatProvider(
             api_key="test", api_base="http://192.168.8.188:1234/v1", spec=spec,
         )
+        await provider._ensure_client()
         pool = provider._client._client._transport._pool
         assert pool._keepalive_expiry == 0
 
-    def test_cloud_keeps_default_keepalive(self):
+    async def test_cloud_keeps_default_keepalive(self):
         spec = _make_spec(is_local=False)
         spec.env_key = ""
         spec.default_api_base = "https://api.openai.com/v1"
         provider = OpenAICompatProvider(
             api_key="test", api_base=None, spec=spec,
         )
+        await provider._ensure_client()
         pool = provider._client._client._transport._pool
         # Default httpx keepalive is 5.0s
         assert pool._keepalive_expiry == 5.0

tests/providers/test_openai_compat_timeout.py

@@ -29,7 +29,7 @@ def test_openai_compat_provider_sets_timeout_on_local_http_client() -> None:
     with (
         patch("nanobot.providers.openai_compat_provider.AsyncOpenAI") as mock_async_openai,
         patch(
-            "nanobot.providers.openai_compat_provider.httpx.AsyncClient",
+            "httpx.AsyncClient",
             return_value=sentinel.http_client,
         ) as mock_http_client,
     ):