Wrong LCAO orbitals when atom is at corner of simulationcell

**Describe the bug**
LCAO wave function not correctly built when an atom is placed at the corner of a simulation cell under open boundary conditions.

**To Reproduce**
Steps to reproduce the behavior:
1. checkout release version 3.10.0
2. build real using intel/19.1.2
```
export CC=mpiicc
export CXX=mpiicpc
cmake
make
```
3. add simulationcell to tests/molecules/O_ae_pyscf_UHF/vmc_short_noj.in.xml
```


20.0 0 0
0 20.0 0
0 0 20.0


n n n


```
4. run tests/molecules/O_ae_pyscf_UHF
```
cd tests/molecules/O_ae_pyscf_UHF/pyscf
python O.py > O.out
cd ..
ln -s pyscf/O-UHF-Triplet.h5
qmcpack vmc_short_noj.in.xml
```

**Expected behavior**
Bogus VMC energy and acceptance ratio
```
$ qmca -q ear vmc_short_noj.s000.scalar.dat
vmc_short_noj series 0
LocalEnergy = -65.13 +/- 0.32
AcceptRatio = 0.13269 +/- 0.00042
```

expected:
```
vmc_short_noj series 0
LocalEnergy = -74.822 +/- 0.036
AcceptRatio = 0.47191 +/- 0.00032
```

**System:**
- Xeon E5-2620 v2
- Fedora 32
- intel/19.1.2 gcc/10.2.1