Reenter KDB on fault on powerpc, instead of panicking

Most other architectures already re-enter KDB on faults, powerpc and mips
are the only outliers.  Correct this for powerpc, so that now bad addresses
can be handled gracefully instead of panicking.

sys/powerpc/powerpc/trap.c

@@ -208,6 +208,13 @@ trap(struct trapframe *frame)
 
 	VM_CNT_INC(v_trap);
 
+#ifdef KDB
+	if (kdb_active) {
+		kdb_reenter();
+		return;
+	}
+#endif
+
 	td = curthread;
 	p = td->td_proc;