[NEMO-devel] improved spatial modelling

[Ben Elliston]

I've just finished making an important change to NEMO that probably
needs to be outlined here.

Previously, NEMO's spatial modelling was based on very large regions
(NEM regions).  Generators were assigned to a region and energy
exchanges were tracked between these regions.  For example, if there is
wind generation in excess of local region demand, that is exchanged with
another nearby region.  This led to a very simple 5-node model for the NEM.

The AEMO 100% RE Study came along with RE resource data for 43 polygons
around the NEM.  These data sets are being used already with multiple
generators using different polygon traces but assigned to the same
region.  It was not possible to properly utilise the higher spatial
resolution of these data.

I've cleaned up the "region" and "polygon" abstractions:

* a region now contains one or more polygons;
* a polygon is defined by a set of coordinates forming a closed polygon;
* a region has one polygon as its major load centre;
* generators are now assigned to a polygon, not a region; and
* a transmission network for the NEM has been encoded and is illustrated
at https://air.net.au/~bje/aemo100.html

The model now represents exchanges between polygons as a 43 x 43 matrix.
 In all other respects, the model works as before, but can now
co-optimise generation mix and transmission costs at "polygon resolution".

Bug fixes to follow over the coming days. :-)

Cheers,
Ben