fix(startup): ping real endpoints in warmup/keepalive (#3641)

_warmup_endpoints called model_discovery.get_endpoints(), which does not exist
on ModelDiscovery. It raised AttributeError on every startup and on every 60s
keepalive tick, was swallowed by the outer except, and pinged nothing, so the
cold-start prevention the loop exists for never ran.

Add ModelDiscovery.warmup_ping_urls(), which resolves the /models probe URLs
from the real discover_models() output, and call it from the warmup loop via
asyncio.to_thread (discovery does a blocking port scan, so keep it off the event
loop).

Adds tests/test_warmup_ping_urls.py: resolves /models URLs from discovered
items, honors the limit, degrades to [] on discovery failure, and documents that
get_endpoints never existed.

tests/test_warmup_ping_urls.py

@@ -0,0 +1,47 @@
+"""Startup warmup must resolve real endpoint URLs.
+
+The warmup/keepalive loop called `model_discovery.get_endpoints()`, which does
+not exist on ModelDiscovery, so it raised AttributeError every run and pinged
+nothing. `ModelDiscovery.warmup_ping_urls()` resolves the /models probe URLs
+from the real discovery API.
+"""
+from src.model_discovery import ModelDiscovery
+
+
+def _md():
+    return ModelDiscovery.__new__(ModelDiscovery)
+
+
+def test_old_method_never_existed():
+    # Documents why the old warmup was a silent no-op.
+    assert not hasattr(ModelDiscovery, "get_endpoints")
+
+
+def test_resolves_models_urls_from_discovered_items():
+    md = _md()
+    md.discover_models = lambda: {"items": [
+        {"url": "http://host:8000/v1/chat/completions", "models": ["a"]},
+        {"url": "http://host:1234/v1/chat/completions", "models": ["b"]},
+    ]}
+    assert md.warmup_ping_urls() == [
+        "http://host:8000/v1/models",
+        "http://host:1234/v1/models",
+    ]
+
+
+def test_limit_caps_results():
+    md = _md()
+    md.discover_models = lambda: {"items": [
+        {"url": f"http://h:{8000 + i}/v1/chat/completions"} for i in range(10)
+    ]}
+    assert len(md.warmup_ping_urls(limit=3)) == 3
+
+
+def test_discovery_failure_degrades_to_empty():
+    md = _md()
+
+    def boom():
+        raise RuntimeError("port scan failed")
+
+    md.discover_models = boom
+    assert md.warmup_ping_urls() == []