Memory issue and RNG not fixed for Gaussian Random Fields.

[odunbar]

I see that repeated runs/trials of darcy.jl and darcy_accelerated.jl leads to randomness (despite attempts to fix RNG) and increasing memory.

[odunbar]

The memory issue was resolved in darcy_accelerated.jl by adding explicit GC calls in between the trials, however, it may be that there is still a way of making memory usage more efficient within an experiment too?