Restoring high degree coordinate fields for icosahedral sphere meshes

[JHopeCollins]

Making a LinearVariationalProblem from the linearisation of a form including the cell normal of a manifold mesh with coordinate degree greater than 1 currently appears to be broken in Firedrake (See issue 2595).

This is causing the shallow water equations in tests/test_paradiag.py on the sphere tests in tests/test_paradiag.pyto break.
The coordinate degree in these tests has been reduced to 1 so that the tests suite can pass, but should be increased again once the upstream fix is in place.

[JHopeCollins]

Gusto have a workaround for this issue here that interpolates the cell normals into a DG Function of sufficiently high order. This Function can then be used in the SWE forms. We could also do this until the issue is fixed in Firedrake.