Update varyiteration variables to use LHS instead of RNG

[ym1906]

I was looking at vary_iteration variables and at the moment it is done in a very rudimentary way, just randomly sampling for each calculation.

We could instead implement LHS sampling for a more efficient and comprehensive sample scheme.

[timothy-nunn]

Seconded. If we wanted to keep the feel of randomness, (shuffled) Sobol could be another good sampler.

[ym1906]

Yes, I have already got a quick implementation working - I'll create a branch. It will be easy for users to change the sampler too. We can also do away with the config file.

That said @jonmaddock mentioned he might like to remove this feature altogether. I think it's still useful to have for quick and simple runs locally.