HEG produces "random" results with batch

HEG calculations seem to produce random but stable results for any particular random seed.

**To Reproduce**
Run a HEG VMC calculation.

```xml

batch




p p p

6
5.0
14



-1
1.0


-1
1.0















1.082858193 0.6653279375 0.4358910287 0.2243616172 0.1102948764




1.696171854 1.047722154 0.6275148566 0.3175982878 0.1446706214









200
100
40
5.0

```

Update the project id to Keep you results.
Run another HEG calculation with the same input.
Compare to the run's scalar.dat's

Run 1
```
# index LocalEnergy LocalEnergy_sq LocalPotential Kinetic ElecElec BlockWeight BlockCPU AcceptRatio
0 5.0510903186e+00 3.8657572467e+01 -6.7322779272e+00 1.1783368246e+01 -6.7322779272e+00 2.5600000000e+04 3.7331729520e+00 5.1770368304e-01
1 5.0135188702e+00 3.8581755845e+01 -6.7352521701e+00 1.1748771040e+01 -6.7352521701e+00 2.5600000000e+04 3.7289637030e+00 5.1763950893e-01
2 5.0632069853e+00 3.9144665819e+01 -6.7381508086e+00 1.1801357794e+01 -6.7381508086e+00 2.5600000000e+04 3.7109413520e+00 5.1728794643e-01
3 5.0037196089e+00 3.8702979247e+01 -6.7330429175e+00 1.1736762526e+01 -6.7330429175e+00 2.5600000000e+04 3.7213199860e+00 5.1833705357e-01
4 5.0543480890e+00 3.8858143057e+01 -6.7358911315e+00 1.1790239221e+01 -6.7358911315e+00 2.5600000000e+04 3.7204689640e+00 5.1724330357e-01
```
Run 2
```
# index LocalEnergy LocalEnergy_sq LocalPotential Kinetic ElecElec BlockWeight BlockCPU AcceptRatio
0 6.6706736285e+00 5.7851923310e+01 -5.1452389250e+00 1.1815912553e+01 -5.1452389250e+00 2.5600000000e+04 3.7391328910e+00 5.1616071429e-01
1 6.6569517588e+00 5.7537601793e+01 -5.1454341345e+00 1.1802385893e+01 -5.1454341345e+00 2.5600000000e+04 3.7358367980e+00 5.1873604911e-01
2 6.6681928152e+00 5.7613986113e+01 -5.1454061489e+00 1.1813598964e+01 -5.1454061489e+00 2.5600000000e+04 3.7326993590e+00 5.1755301339e-01
3 6.6648498744e+00 5.8741626524e+01 -5.1429758167e+00 1.1807825691e+01 -5.1429758167e+00 2.5600000000e+04 3.7283851180e+00 5.1908203125e-01
4 6.6542603439e+00 5.7664675633e+01 -5.1458908133e+00 1.1800151157e+01 -5.1458908133e+00 2.5600000000e+04 3.7241059530e+00 5.1715959821e-01
```
Note: I did run these two with a tau of 10 instead of 5 in the hope of breaking out of a minima (not that that makes sense to me) tau=5 gives a higher acceptance ratio but behavior is otherwise the same.

**Expected behavior**
Block energies that converge within error bars.

**System:**
- sdgx-server

**Additional context**
I was hoping to use heg to establish correctness for the batched Energy Density Estimator, but obviously if it can't converge...