fix(monitors): a failed ping or TCP check says why (#5430)

tcp_check swallowed every exception and icmp_check returned on a
non-zero exit without a reason, so every down ICMP/TCP monitor stored
error_msg NULL and the Monitors page and MCP showed it down with no why.

Each failure now names its cause: "No reply within 2 s", "Connection
refused on port N", "No answer on port N within 5 s", "Can't resolve
<name>: ...", the OS error, or ping's own first stderr line. An ICMP
check that fell back to TCP says so, and a ping that outlives its
timeout is killed rather than left running.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/core/test_ping_checks.py

@@ -0,0 +1,65 @@
+"""A failed ping or TCP check says why (#5430): no reply, refused, no answer
+in time, a name that doesn't resolve, or ping's own complaint; an ICMP check
+that fell back to TCP says so."""
+import asyncio
+import socket
+
+import pytest
+
+from steward.monitors import ping
+
+
+@pytest.mark.parametrize("exc,want", [
+    (asyncio.TimeoutError(), "No answer on port 22 within 5 s"),
+    (ConnectionRefusedError(111, "Connection refused"), "Connection refused on port 22"),
+    (socket.gaierror(-2, "Name or service not known"), "Can't resolve nas.lan: Name or service not known"),
+    (OSError(113, "No route to host"), "Can't connect to port 22: No route to host"),
+    (RuntimeError("odd"), "odd"),
+])
+def test_a_tcp_failure_says_why(monkeypatch, exc, want):
+    async def boom(*a, **kw):
+        raise exc
+    monkeypatch.setattr(ping.asyncio, "open_connection", boom)
+    got = asyncio.run(ping.tcp_check("nas.lan", 22))
+    assert got == {"is_up": False, "response_ms": None, "error_msg": want}
+
+
+class _Proc:
+    def __init__(self, returncode, stderr=b""):
+        self.returncode, self._stderr = returncode, stderr
+
+    async def communicate(self):
+        return b"", self._stderr
+
+
+@pytest.mark.parametrize("code,stderr,want", [
+    (1, b"", "No reply within 2 s"),
+    (2, b"ping: nas.lan: Name or service not known\n", "ping: nas.lan: Name or service not known"),
+    (2, b"", "ping failed (exit 2)"),
+])
+def test_a_ping_failure_says_why(monkeypatch, code, stderr, want):
+    async def spawn(*a, **kw):
+        return _Proc(code, stderr)
+    monkeypatch.setattr(ping.asyncio, "create_subprocess_exec", spawn)
+    got = asyncio.run(ping.icmp_check("nas.lan"))
+    assert got == {"is_up": False, "response_ms": None, "error_msg": want}
+
+
+def test_a_reply_is_up_with_no_error(monkeypatch):
+    async def spawn(*a, **kw):
+        return _Proc(0)
+    monkeypatch.setattr(ping.asyncio, "create_subprocess_exec", spawn)
+    got = asyncio.run(ping.icmp_check("nas.lan"))
+    assert got["is_up"] is True and got["error_msg"] is None
+
+
+def test_without_ping_the_tcp_stand_in_says_so(monkeypatch):
+    async def no_ping(*a, **kw):
+        raise FileNotFoundError("ping")
+
+    async def refused(*a, **kw):
+        raise ConnectionRefusedError(111, "Connection refused")
+    monkeypatch.setattr(ping.asyncio, "create_subprocess_exec", no_ping)
+    monkeypatch.setattr(ping.asyncio, "open_connection", refused)
+    got = asyncio.run(ping.icmp_check("nas.lan"))
+    assert got["error_msg"] == "ping unavailable, TCP check instead: Connection refused on port 80"