Fix CI errors

[cpjordan]

Fix various issues with testing:

• ufl_element() -> ufl_function_space()
• Fatal Python error: Aborted - TO BE FIXED
• AttributeError: 'Cofunction' object has no attribute 'ufl_class' - TO BE FIXED

[stephankramer]

Thanks @cpjordan
So it might be I'm just misunderstanding - but I would have expected that for interpolated value size we should in fact be using value_size and not block_size. The latter being the n/o DOFs per node, which was what previously returned by value_size in which case it would in fact have been incorrect in Thetis previously. The difference would only appear for RT and BDM velocity fields which we don't use very frequently. I do agree/understand the changes to the OP2 kernel headers in utility.py and utility3d.py - these should indeed be block_size now.

[stephankramer]

Related to firedrakeproject/firedrake#3862 and firedrakeproject/firedrake#3368

[cpjordan]

@stephankramer - thanks for reviewing and explanation. I had discussed the problems with @connorjward who found that value_size in the utility kernels was the cause of the failing tests, but I just replaced all instances assuming that the previous implementation in Thetis was correct - I should have looked into it more!

Updated based on the comments above.

[stephankramer]

Perfect, many thanks!