[patch] new dg-effective-target(s) for >= 32 bits, plus testcase patches

[DJ Delorie]

> Do you mean for tests of functionality that ought to be supported
> for the target?

No, for things that aren't supported by the target, when lack of
support causes cascading failures.

> A test should be skipped if the functionality is not supported for
> the target, or xfail'ed if it demonstrates a known bug.

I must be out of touch then.  I thought XFAIL was for "we expect this
test to fail because the target can't support it".  Are we now using
UNSUPPORTED for this?