Abstract methods for creating catchment indicators

[carlhiggs]

Is your feature request related to a problem? Please describe.
We currently have a method for creating catchment indicators that is more or less hard coded for population density and intersection density per sqkm within 1000m walkable neighbourhood catchments. We use an approximation method to make city-scale analysis for sample points feasible:

1. preprocessing a population grid (eg 100m resolution) with relevant attributes (e.g. count or density)
2. evaluate neighbourhood statistics for nodes, by identifying the grid cells reachable within a specified distance (ie. 1000m)
3. calculate proximity-weighted average of density statistics for sample points

Describe the solution you'd like
The current implementation should be abstracted and made more general to support adding new kinds of catchment indicators (density, following the same pattern, but potentially other kinds too). This ties in with issue #343.

Potentially we should also investigate feasibility of calculating statistics for all sample points directly; given other improvements in process efficiency, that may be feasible now. The challenge remains managing time and memory intensive processing, particularly given uncertainty of computational resources available on users' computers - might be laptop, might be powerful workstation. That would have been the historic rationale for the approximation approach - but always good to revisit these things.