Bugfix for worker creation with small execution time

When running dummy scripts the sem_timedwait was unable to detect that the
pool was busy...

src/responder.c

 	pid_t ret;
 	/**@brief poll timeout */
 	struct timespec timeout;
+	struct timespec idle_timeout;
 	/**@brief watchdog timeout */
 	struct timespec pool_timeout;
 	time_t idle_start, busy_start;
 		exit(PYFCGI_FATAL);
 	}
 
+	idle_timeout.tv_sec = 0;
+	idle_timeout.tv_nsec = 100000; //0.0001s
 	timeout.tv_sec = 0;
-	timeout.tv_nsec = 100000000;
+	timeout.tv_nsec = 100000000; //0.1s
 	idle = busy = 0;
 
 	pyfcgi_logger_set_ident("Workpool");
 		PyFCGI_conf.context.n_wrk = n_wrk;
 	}
 	//Wait at least for a process to be ready
-	while(!pyfcgi_pool_idle(&timeout));
+	while(!pyfcgi_pool_idle(&idle_timeout));
 
 	last_update = 0;
 
 			// respawn on same slot
 			pyfcgi_log(LOG_DEBUG, "respawning worker #%d", n);
 			wrk_pids[n] = spawn(n);
-
-			continue;
 		}
 
 		// Check if the pool is idle or busy
-		if(pyfcgi_pool_idle(&timeout))
+		if(pyfcgi_pool_idle(&idle_timeout))
 		{
 			// workers idle
 			busy = 0;