allow GlmSpeedCallback without StepsizeCallback

[jlchan]

Is there a good way to use RDPK3SpFSAL49() with the existing MHD solvers? A naive swap results in errors.

For example, running

trixi_include("examples/unstructured_2d_dgsem/elixir_mhd_ec.jl", 
              sol = solve(ode, RDPK3SpFSAL49(), abstol=1e-7, reltol=1e-7, save_everystep=false, callback=callbacks))

prints the errors

┌ Warning: First function call produced NaNs. Exiting.
└ @ OrdinaryDiffEq ~/.julia/packages/OrdinaryDiffEq/9xnA1/src/initdt.jl:108
┌ Warning: Automatic dt set the starting dt as NaN, causing instability.
└ @ OrdinaryDiffEq ~/.julia/packages/OrdinaryDiffEq/9xnA1/src/solve.jl:510
┌ Warning: NaN dt detected. Likely a NaN value in the state, parameters, or derivative value caused this outcome.
└ @ SciMLBase ~/.julia/packages/SciMLBase/UIp7W/src/integrator_interface.jl:325

I think this is due to the fact that equations.c_h = NaN on initialization, and needs GlmSpeedCallback to be set properly. This seems to be done when using CarpenterKennedy2N54 but not RDPK3SpFSAL49().

[ranocha]

The GLM MHD equations use a pretty hacky way of setting the GLM speed: They assume that you run with a CFL callback and use that to update the GLM speed after every step. For now, you can only disable adaptive time stepping or initialize the GLM speed with some constant value to use throughout the simulation... To make this work more cleanly, we need to update the GLM callback.