Rckirby/explicit

[rckirby]

This PR does two things:

1. It adds an optimized explicit time stepper (which inherits from DIRK since it's an SDIRK). One accesses this with stage_type="explicit" instead of stage_type="dirk". It automatically sets up SNES to never rebuild the jacobian or preconditioner, which is an optimization that works for RK. So users just have to provide mass matrix inversion parameters.
2. It also adds a new family of "pseudo-energy-conserving" explicit RK methods from de Leon, Ketcheson, and Ranocha.

[ScottMacLachlan]

I would love to see a test or two focused on the explicit time steppers. Is there a reason to not expand test_wave_energy.py to run these schemes through their paces?

[rckirby]

I would love to see a test or two focused on the explicit time steppers. Is there a reason to not expand test_wave_energy.py to run these schemes through their paces?

Yes, good point. I've done this now.