Ion-electron collisions in Boltzmann model & option to run with fixed fraction

[manauref]

(see issue #967 for more discussion and future plans)

In this work we add ion-electron collisions to the Boltzmann electron model. In order to support spatially and time-dependent collision frequency $\nu_{ie}=\nu_{ie}(x,t)$ we:

1. Need the electron density to be updated in space-time. This is accomplished by including an electron species in the Boltzmann electron input file, but without any terms, and adding a scaling operation:

.elc = {
  .lower = ...,
  .upper = ...,
  .cells = ...,

  .scaling = {
    .type = GKYL_GK_SPECIES_SCALING_BOLTZMANN,
  },
},

which simply sets, every time step, the electron density as

$$n_e = n_{i,sheath} \exp\left[-e\left(\phi - \phi_{sheath}\right)/T_e\right]$$

assuming quasineutrality at the sheath entrance.

2. Need the electron collisions app to be initialized (because ions get $nu_{ei}$ from it), so we add the electron collision table but we add the option do_not_add_to_dfdt so that electrons are not updated by electron-ion collisions. That is, electron table in the input file would look like:

.elc = {
  .lower = ...,
  .upper = ...,
  .cells = ...,

  .scaling = {
    .type = GKYL_GK_SPECIES_SCALING_BOLTZMANN,
  },

  .collisions = {
    ...
    .do_not_add_to_dfdt = true,
  },
},

New fixed fraction

We upgraded the previously existing .recycle_react_scaling operation that we implemented for scaling neutrals to simply .scaling, and so we can scale the electrons as mentioned before. In the process we also added an option to scale any species to a fixed fraction of another species. For example, we may want to run with a fixed fraction carbon species like

.carbon = {
  .scaling = {
    .type = GKYL_GK_SPECIES_SCALING_FIXED_FRACTION,
    .reference_species = 'ion',
    .fixed_fraction = 0.05,
  },
},

so that every time step it sets the carbon density to be 5% of the main ion density.

Tests

@Maxwell-Rosen ran mirror simulations with and w/o ion-elc collisions, and below we can see that they have the intended effect:

• Lower density because we increase the rate at which we scatter into the loss cone
• Lower temperature because beam ions are more quickly slowed down

We also ran all the GK regression tests with the new runregression tool @JunoRavin implemented. They all pass, except for rt_gk_multib_sheath_1x2v_p1, but that should be unaffected by this work and I think there's an issue in the way that test is constructed.