NetCDF outputs for 2D fields

[tkarna]

Adds netCDF outputs for 2d fields. Usage via the new option:

options.fields_to_export_netcdf = ['uv_2d', 'elev_2d']

Supports:

• 2D fields (also in the 3D model)
• P0, P1, and P1DG function spaces
  • all other functions will be interpolated to P1DG space prior to exporting
• triangular and quad meshes
• scalar and vector fields
• output only (netcdf files cannot be read back in)

Benefits:

• portable output format (unlike vtk)
• smaller files (mesh stored only once)

Outputs will be stored in files outputs/nc/Elevation2d/Elevation2d.nc. All time steps will be appended in the same file. NetCDF header looks like this

netcdf Elevation2d {
dimensions:
        face = 800 ;
        face_nb_nodes = 3 ;
        vertex = 2400 ;
        time = UNLIMITED ; // (99 currently)
variables:
        uint64 Mesh_face_nodes(face, face_nb_nodes) ;
                Mesh_face_nodes:start_index = 0LL ;
        float Mesh_node_x(vertex) ;
        float Mesh_node_y(vertex) ;
        float time(time) ;
                time:standard_name = "time" ;
                time:units = "s" ;
        float elev_2d(time, vertex) ;
}

The stored mesh and data are defined as follows:

P0 fields:

• connectivity and mesh coordinates: P1
• data: P0 (a.k.a. "cells")

P1 fields:

• connectivity and mesh coordinates: P1
• data: P1 (a.k.a. "vertices")

P1DG fields:

• connectivity and mesh coordinates: P1DG
• data: P1DG (a.k.a. "vertices")

To enable netcdf outputs firedrake must be installed with:
PETSC_CONFIGURE_OPTIONS="--download-netcdf-configure-arguments=--enable-parallel"
In addition, netCDF4-python package has to be installed manually because its default configuration fails to detect parallel netcdf4 capability. setup.py needs to be modified to set has_parallel4_support = True.