crypt32: Fix a test failure on systems with no c: drive

[Juan Lang]

> Wouldn't it be cleaner to change the use of the hardcoded c:\deadbeef.dbf?

I don't think so.  The point is we expect the file not to be there,
and we're testing what the resulting GUID is when the file doesn't
exist.  We're less concerned about the particular last error.  I
suppose I could have removed it.  On the other hand, if for some
strange reason that file existed, we might see a different error, so
we'd at least know that the failing test is due to the file existing,
rather than a change in the behavior of wintrust.

So, long story short:  I like my patch just fine ;)
--Juan