Realistic household age-distributions

[ptheywood]

The initial random implementation does not aim to produce households of realistic age distributions, each household is generated with members of random age's (based on age demographic paramaters), leading to unreaslistic households (i.e. 6 0-9 year olds).

In the reference model, rejection sampling is used when generating a new individual and assigning them to a household, only accpeting the new household if it does not degrade the error (sum of square difference) beyond a threshold.
The refernce household demographics are read from a file (from a cli argument). The default reference file is 10000 households, with the number of individuals in each age band (a_0_9,a_10_19,...)