Continue holding lock while finding error location SERVER-5101

This has no impact on the normal case since we return anyway. Current
test fails are indicating that the problem was written with
writeNoJournal, but that looks correct. This should confirm or rule that
out.

src/mongo/db/dur.cpp

@@ -359,11 +359,10 @@ namespace mongo {
                     _bytes += mmf->length();
 
                     assert( mmf->length() == (unsigned) mmf->length() );
-                    {
-                        scoped_lock lk( privateViews._mutex() ); // see setNoJournal
-                        if (memcmp(p, w, (unsigned) mmf->length()) == 0)
-                            return; // next file
-                    }
+
+                    scoped_lock lk( privateViews._mutex() ); // see setNoJournal
+                    if (memcmp(p, w, (unsigned) mmf->length()) == 0)
+                        return; // next file
 
                     unsigned low = 0xffffffff;
                     unsigned high = 0;