Partially revert the previous change. I failed to notice that where

ktruserret() is invoked, an unlocked check of  the per-process queue
is performed inline, thus, we don't lock the ktrace_sx on every userret().

Pointy hat to:	jhb
Approved by:	re (kensmith)
Pointy hat recovered from:	rwatson

sys/kern/kern_ktrace.c

@@ -444,8 +444,6 @@ void
 ktruserret(struct thread *td)
 {
 
-	if (STAILQ_EMPTY(&td->td_proc->p_ktr))
-		return;
 	ktrace_enter(td);
 	sx_xlock(&ktrace_sx);
 	ktr_drain(td);