DNS_Mgr: Use Process() for timeout expiration

DNS_Mgr has a GetNextTimeout() implementation that may return 0.0. When
that is the case, its IO source is enqueued as ready with an fd of -1.
This in turn results in Process() being called instead of ProcessFd()
in RunState.cc.

Ensure timeouts behavior is properly handled by actually forwarding
timeout indications to c-ares via DNS_Mgr::Process(). This results
in pending DNS queries for which a timeout happened to actually
timeout (when there's no other connectivity).

src/DNS_Mgr.cc

@@ -1471,6 +1471,15 @@ void DNS_Mgr::ProcessFd(int fd, int flags)
 	IssueAsyncRequests();
 	}
 
+void DNS_Mgr::Process()
+	{
+	// Process() is called when DNS_Mgr is found "ready" when its
+	// GetNextTimeout() returns 0.0, but there's no active FD.
+	//
+	// Kick off timeouts at least.
+	ares_process_fd(channel, ARES_SOCKET_BAD, ARES_SOCKET_BAD);
+	}
+
 void DNS_Mgr::GetStats(Stats* stats)
 	{
 	// TODO: can this use the telemetry framework?

src/DNS_Mgr.h

@@ -285,7 +285,7 @@ protected:
 	void IssueAsyncRequests();
 
 	// IOSource interface.
-	void Process() override { }
+	void Process() override;
 	void ProcessFd(int fd, int flags) override;
 	void InitSource() override;
 	const char* Tag() override { return "DNS_Mgr"; }