Fix bogus active counting for procs terminating due to down-scaling

Be sure to mark the proc as idel before actually doing "exit()", otherwise we will count it as busy for the whole time the proc is terminated by OS (we automatically clear the load stats only upon sigchild on attendant)

(cherry picked from commit e982e90)

pt.c

@@ -486,6 +486,8 @@ void dynamic_process_final_exit(void)
 	/* if a TCP proc by chance, reset the tcp-related data */
 	tcp_reset_worker_slot();
 
+	pt_become_idle();
+
 	/* mark myself as DYNAMIC (just in case) to have an err-less termination */
 	pt[process_no].flags |= OSS_PROC_SELFEXIT;
 	LM_INFO("doing self termination\n");