Gusto performance improvements

[JHopeCollins]

Some places where performance could potentially be improved - to discuss at hackathon?

1. Matrix refactorisations Implemented in PR Prevent unnecessary refactorisations when matrices haven't changed #569
2. Using Cofunction for SIQN linear solver right hand side to avoid the mass solve in the forcing calculation entirely.
3. Finding alternative to SCPC for the hybridisation solves in the SIQN linear solvers to avoid refactoring local solves every iteration.
4. SSPRK schemes with higher stage count and higher effective CFL (CFL limit/# stages). E.g. 3rd order 4 stage has CFL limit 2 (effective CFL = 0.5), whereas 3rd order 3 stage only has CFL limit 1 (effective CFL = 0.33). Implemented in Add SSPRK schemes with higher stage count #606