Sudden jump in binding energy with separation distance

**Describe the bug**
Binding energy of 3x3x1 bilayer graphene jumps by 0.6 eV/atom, suddenly at interlayer separation of 5.0 A.

This behavior is shown below for VMC (no jastrow, PBE orbitals):


By also performing calculations at 4.99 and 5.01 A, the source of the jump can be isolated to the e-e interaction. Units below are eV/atom.

| Component | E(5.0) | (E(4.99)+E(5.01))/2 | Difference |
| ----------- | ----------- | ----------- | ----------- |
| Total energy | -149.10(2) | -149.79(3) | **0.69(3)** |
| Kinetic | 117.80(6) | 117.77(9) | 0.03(9) |
| Ion-Ion | 1169.7759 | 1169.7769| 0.001 |
| Local ECP | -2701.1(2) | -2701.1(3) | 0.3(3) |
| Non-local ECP | 11.09(4) | 11.08(6) | 0.01(8) |
| Elec-Elec | 1253.4(1) | 1252.7(2) | **0.7(2)** |

Error shows insensitivity to the variations we have tried:
- Use of hybrid rep
- Lr_dim cutoff (25-100)

The error also appears consistently at every statistical block. Plot below shows averages over 10 blocks for clarity. Therefore, the error could be more likely to reside in the construction of the potential, rather than a feature of the configuration sampled during the random walk.


DFT curve is smooth at these interlayer separations and occupations are either 0 or 1 down to machine precision.

**To Reproduce**
Steps to reproduce the behavior:
1. release version or git commit hash being built :
Git branch: develop
Last git commit: 89af1d7f49ee536b8046cce24da561858da10bc9-dirty
Last git commit date: Fri Nov 4 10:37:03 2022 -0400
Last git commit subject: Merge pull request #4304 from markdewing/orb_rot_hcpBe
2. cmake command: Andes build script

Files needed to reproduce are located at: /gpfs/alpine/mat151/proj-shared/qmcpack_issue_4549
All the results are from a complex build

**Expected behavior**
Binding curve should be smooth.

**System:**
- system name: Andes