bugfix out-of-bound err HP - 2nd attempt

Changed again the test inputfile - should work now. However, the error affects this calculation also in the collinear, non-relativistic version. Indeed, it might connected to the fact that the internal 3x3x3 mesh used for +U+V calculation is independent from the q-mesh used in linear response. In the post-processing of the Hubbard parameters, when the two meshes interact, such error arises. It could likely be that this test is an unrealistic corner case though: no-one would use a 1x1x1 q-mesh for BN.