ole32: Accept STG_E_UNIMPLEMENTEDFUNCTION when ILockBytes implementation doesn't support locking.

[Vincent Povirk]

This looks OK to me, and if it fixes a regression I don't think we
need a test case (though it never hurts).

I don't think StorageImpl_LockRegionSync should be filtering these
codes (the caller needs to know the difference between a failure and
unimplemented function), though we should probably also fix the call
inside StorageImpl_LockTransaction, the check in UnlockTransaction,
and the check in StorageImpl_LockRegionSync. And it seems we also
check for E_NOTIMPL when we call LockTransaction..

Maybe an HR_IS_UNIMPLEMENTED macro would be warranted?