NHE: mesh object and rupture-surface-mesh interaction refactoring

[Anton Gritsay]

That's what should be done:

1) have a class Mesh meant to represent in a efficient way a mesh of locations (that is by using numpy arrays to store separately longitude, latitude, and depths of the points representing the mesh). It supports a method for calculating mesh centroid (that can be used for setting the hypocenter in fault ruptures), and also support methods for distance calculation:
       - get_mesh_centroid
       - get_minimum_distance(location)

2) have and abstract class BaseSurface meant to represent any surface. It defines the following methods:

       - get_surface_mesh(), which returns a mesh representation of the surface
       - get_strike(), compute strike [azimuth] of the surface
       - get_dip(), compute dip [inclination] of the surface
       - get_minimum_distance(location) [Calls the method get_surface_mesh() and call the method get_minimum_distance() of the corresponding mesh object]
       - get_other_distances ...
The abstract class provide an implementation of the get distance methods based on the surface mesh representation. Each class implementing BaseSurface can then implement a more optimized version if it's possible.

3) change planar surface object so it signature looks like this:

       - PlanarSurface([corners],strike,dip,mesh_spacing)
       It's defined by a list of 4 cornes, strike, dip, and mesh_spacing (to be used to construct the mesh). It's used to define ruptures from PointSource and AreaSource. The methods get_strike and get_dip simply returns the strike and dip values given in the constructor.