Channel-Benchmark: CoreNeuron+MOD2C (GPU execution) - Different Spikes compared to NRN & weird messages regarding random number generation #593

To reproduce this issue:
1. Clone channel benchmark (from bbpgitlab) and checkout to update_scripts branch.
2. Run the install script.
3. Submit (sbatch) the run_gpu.sh script.
4. The spikes generated from NRN simulation (NRN_GPU.spk) are a bit different from the ones generated from CoreNeuron+MOD2C (GPU_MOD2C.spk). The spikes can be found in benchmark/channels folder.
5. Inspect the GPU_MOD2C.log (in benchmark/channels folder) for some weird messages regarding random numbers that are generated:
Failure! 1.000000 at time 0.114331: urand = -nan Failure! 1.000000 at time 0.111089: urand = 2.84441e-304 psolve | | t: 0.00 ETA: 0h00m00s Failure! 1.000000 at time 0.114534: urand = -61.9845 Failure! 1.000000 at time 0.134616: urand = 2.35799e-310 Failure! 1.000000 at time 0.123452: urand = -0.000112443 Failure! 1.000000 at time 0.112969: urand = 1.29568e-304 || SYN_ID: 1.000000, release failure || Failure! 1.000000 at time 0.158528: urand = 2.35795e-310 || SYN_ID: 1.000000, release failure || Failure! 1.000000 at time 0.154727: urand = 2.358e-310 Failure! 1.000000 at time 0.170536: urand = -0.000296501 Failure! 1.000000 at time 0.179273: urand = 2.35799e-310 Failure! 1.000000 at time 0.173388: urand = 1.78944e-309 Failure! 1.000000 at time 0.178487: urand = 2.35795e-310 Failure! 1.000000 at time 0.21024: urand = -0.000292664 Failure! 1.000000 at time 0.179791: urand = -0.000405227 || SYN_ID: 1.000000, release failure || || SYN_ID: 1.000000, release failure || Failure! 1.000000 at time 0.198301: urand = 8.12061e-06 Failure! 1.000000 at time 0.191264: urand = 9.31414e-15 ...