Revert new experimental libev backend due to time constraints

Due to time constraints I had to revert the new libev backend to the
service, I/O and signals monitor.  Employed instead is the tried and
tested signals based service monitor, with a small extension for an
additional I/O monitor using poll() for the I/O plugins to use.

Tried restoring the startx finit.conf directive on top of the service
implementation. However, it's currently not usable to run /usr/bin/startx
for any other user than root. I recommend using the examples/finit.conf
and run /usr/sbin/gdm instead.

finit.c:
        New run_loop() function, replaces libev ev_run().

helpers.c:
        Refactor start_process() into svc_start().

plugin.c:
        Add fds[] array for use in the io_monitor() called by the
        new run_loop().

signal.c:
        Ignore SIGCHLD. Handled by waitpid() in svc_monitor() now,
        which is called by the new run_loop().

svc.c:
        svc_register(): Optional @username parsing for the command line.
        svc_monitor(): Replaced libev callback with waitpid()
        svc_start(): Refactored to use most of helpers.c:start_process()

Signed-off-by: Joachim Nilsson <troglobit@gmail.com>

conf.c

@@ -122,9 +122,12 @@ void parse_finit_conf(char *file)
 				network = build_cmd(NULL, x, CMD_SIZE);
 				continue;
 			}
+			if (MATCH_CMD(line, "startx ", x)) {
+				svc_register(x, username);
+				continue;
+			}
 			if (MATCH_CMD(line, "service ", x)) {
-				if (svc_register(x))
-					_e("Failed, too many services to monitor.\n");
+				svc_register(x, NULL);
 				continue;
 			}
 		}