Migrate to RKPlus #37
Description
This is probably something I should have mentioned months ago, but I rewrote a number of the (explicit) time integrators and tested them thoroughly in the RKPlus sctipt. The thinking behind this was that RK2 was being used as the default integrator a lot of the time, but most models require the use of RKSplit due to their source terms. The integrators in RKPlus all combine the source terms and flux terms in the RHS, rather than strang splitting approaches, and so also allow for higher order time integration with non-stiff sources. Theres the standard RK2, RK3 and two RK4s (one with ten stages that supposedly allows you to break the CFL constraint).
I was going to deprecate the use of the RK2, RKSplit and RKSplit2 integrators in favour of RK2B (or just removing them entirely and renaming RK2B as RK2), but many things got in the way! This shoudn't take too long so if it hasnt been done in a while I'll try to find some time to do it, although I dont want it to clash with anything you (Ania) are doing, as you seem to be working on many things at once. There is also the chance that it might mess up some of your GPU/CPU comparison (maybe you should keep RK2/Split for this reason), and obviously the RKPlus integrators havent been written to run on GPUs.
Let me know what you think and if it will pose any problems.