Increase atol in test_qgt_itersolve.py::test_qgt_dense

Created by: wdphy16

As discussed here

I think atol=1e-17 is too small for float64, while atol=1e-15 is already used here.