Allow close but non-identical noise evidence values when comparing for consistent data

Currently when combining results with a ResultList is checks that the log noise evidences are the same for each result. While this should generally be true, if the values have been calculated on a slightly different CPU core architecture (as might be the case on the computer cluster with different cores) then there might be minor numerical precision differences. This slightly relaxes the check that values must be identical and instead used np.allclose to make sure the absolute difference between values is no greater than 1e-8.