Warn of possibly insufficient quadrature degree for integral on higher-order geometry

[Jack Hale]

Consider the following integral on a non-affine mesh, e.g. a quadratic triangular mesh:

f = assemble(1*dx(domain=mesh, degree=1))

The degree of the quadrature is insufficient to integrate the area of the mesh. Suggest implementing a soft warning to user that they might be doing something silly.

See discussion here:

https://fenicsproject.slackarchive.io/quadhexmesh/page-2

[Miklós Homolya]

Related: https://bitbucket.org/fenics-project/ufl/issues/80

Not sure if FFC is the right place to warn about it, or this should rather done in UFL.

[Jack Hale]

See also: https://bitbucket.org/fenics-project/ufl/issues/80

@blechta suggestion was to place this in as a stopgap while consensus is made on UFL fix.

Already mentioned by @miklos1 above.

[Jan Blechta]

This needs to be done after/in compute_form_data. Do you think that it would be good to blow the function even more? Anyway, I will check what looks better/easier then.