val_dname_bogus: give up on query-minimization

For now at least. This basically reverts 68abde7f which tried to make it work, but this negative trust anchor was a really bad idea. The test (also) relies on names underneath getting validated and used by aggressive cache later. (They are also signed as belonging to the root zone, so it's ugly.)

With "previous" kresd versions the problematic NSEC inside somehow validated as secure, but I don't want to be restricted here.