cbus: use a more reliable test to check whether we need a syscall

Check ev_async_pending() rather than state of the pipe to find
out whether we need a syscall. On a simple read-only test pipe->pipe
is empty while there is already a pending event more than a few times a
millisecond.

At the same time, checking ev_async_pending() is more expensive, since
it's an atomic variable, but overall performance increased
after the patch, so the cost seems to be justified.

src/cbus.cc

@@ -169,7 +169,7 @@ cbus_flush_cb(ev_loop * /* loop */, struct ev_async *watcher,
 	bool peer_output_was_empty = STAILQ_EMPTY(&peer->output);
 
 	cbus_lock(pipe->bus);
-	pipe_was_empty = STAILQ_EMPTY(&pipe->pipe);
+	pipe_was_empty = !ev_async_pending(&pipe->fetch_output);
 	/** Flush input */
 	STAILQ_CONCAT(&pipe->pipe, &pipe->input);
 	/*
@@ -201,7 +201,7 @@ cpipe_peek_impl(struct cpipe *pipe)
 	cbus_lock(pipe->bus);
 	STAILQ_CONCAT(&pipe->output, &pipe->pipe);
 	if (! STAILQ_EMPTY(&peer->input)) {
-		peer_pipe_was_empty = STAILQ_EMPTY(&peer->pipe);
+		peer_pipe_was_empty = !ev_async_pending(&peer->fetch_output);
 		STAILQ_CONCAT(&peer->pipe, &peer->input);
 	}
 	cbus_unlock(pipe->bus);