Noting here for future interest that there's been a bug in PySCF with the Becke grid
pyscf/pyscf#2490
pyscf/pyscf#2499
and it's probably not a bad idea to double check that our quadrature approaches are the same, even though I am sure ours are correct.