Spherical Quad/Triangle orientation

Spherical quads are now oriented in the gnomonic plane according to the first 2 edges

the determinant of the jacobian has to be positive, for correct elements, in 2d

also, add a test for the spherical quad, based on a coupling case between a made-up homme field and mpas mesh

coupling test is launched with something like:

mpiexec -np 2 ./mbcoupler_test -meshes sphere_16p.h5m mpas_p8.h5m -itag vertex_field -meth 4 -outfile out.h5m

A vertex field tag is transferred from the cube sphere mesh to an mpas mesh, using interpolation; no conservation is checked

A spherical triangle element is added too; orientation is again decided by the first 2 edges; interpolation, transfer happens again in the gnomonic plane; which used to be a problem for kdtrees, because the relatively high curvature for coarse spherical triangles ; Because of this issue, a bounding box method was added for spherical elements; Basically, curvature of the sphere (edges) is taken into consideration, so tolerances do not have to be increased artificially anymore

One way to test coupling with spherical triangles:

mpiexec -np 2 ./mbcoupler_test -meshes tri_fl_8p.h5m mpas_p8.h5m -itag vertex_field -meth 4 -outfile oo.h5m -eps 1.e-9

in which tri_fl_8p.h5m file is a partitioned triangle mesh on a sphere, with a made-up vertex field, that will be transferred to an mpas mesh;