Modify the test cases (there are a bunch of them) under test/types so that they

employ different executable names when creating the target for lldb to debug.
Comparing the log files for the same test case with success/failure results show
that the failed one was setting the breakpoint on some different address than the
succeeded one, leading us to suspect that the file system maybe think it knows better
and caches the wrong executable file (they were all named 'a.out' before) which lldb
subsequently reads from.

Now './dotest.py -v types' passes without specifying the '-w' option.

rdar://problem/9644488
rdar://problem/9649573

llvm-svn: 133649