Randomisations - add option to use a spatial condition to calculate the target richness of each group

[shawnlaffan]

Following from the thought in #594, it might be useful to allow a spatial condition to define the target richness of each group in a structured randomisation. That way one could set a target to be the richness of a circular radius around each location, for example.

This is distinct from setting a target richness for a neighbourhood, as adjusting and tracking richness scores for overlapping neighbourhoods would be very complex to support (need to track effect of allocation to a group on all groups for which it is a neighbour). Non-overlapping neighbourhoods might be more feasible, though.