Bad elec-elec values for offload builds when energy density activated.

QMCPACK built with openmp offload and run in batched mode when the energy density operator is activated, elec-elec interactions are not correctly updated. This resulted in bad values in the `scalar.dat` output as well as the `stat.h5`

The issue here turns out to be in `CoulombPBCAA` in the QMCHamiltonian model. This is at the heart of #5758 and that same issue occurs for spline based wavefunctions as well.

**To Reproduce**
see #5758 but run example like diamond 1x1x1.

**Expected behavior**
No affect on elec-elec full system energy values due to enabling an estimator.
Correct per particle energies for elec-elec interactions.

**System:**
- any
-
**Additional context**
Add any other context about the problem here.