Although G=0 is not divergent for HSE, the large changes near the
origin cause problems in the numerical calculation of the nonlocal
exchange energy. To correct this a "divergence" correction analogously
to PBE0 is introduced.

subtract exp(-rcut^2*G^2)*V(G) to make Fourier transform stable near G=0
set all energy contributions of G=0 in main loop to 0
add divergence correction at the end modifying the PBE0 one appropriately

new function in HSEFunctional is used to calculate the scaling of the
HSE "divergence" relative to the PBE0 one
InteractionPotential.h contains this additional function

the integrals ~k^2 are not evaluated because they are of higher order
(in PBE0 the potential is ~k^{-2} near G=0, so that the k^2 terms are
important, in HSE the potential is ~const. near G=0, so that the k^2
terms do not contribute a lot

git-svn-id: http://qboxcode.org/svn/qb/branches/hse-dev@1404 cba15fb0-1239-40c8-b417-11db7ca47a34