fix the Jacobian for the simplified-SDC system

[zingale]

For simplified SDC we need to convert from the reaction system Jacobian (in terms of X, T, e) to the system for the conserved variables. This does that transformation.

[zingale]

this will close #147

still need to do the MAESTROeX / enthalpy part

[zingale]

This now works for both the energy and enthalpy-based SDC formulations.

[zingale]

The energy based Jacobian derivation is here:

simplified_sdc_derivations.pdf

The enthalpy based Jacobian derivation is here:

maestroex_simplified_sdc_derivations.pdf

[zingale]

For the test_sdc unit test, with these changes, when we run with the numerical Jacobian, we get:

Run time = 19.73930645
min number of rhs calls: 21
avg number of rhs calls: 445
max number of rhs calls: 6883

and with the analytic Jacobian we get:

Run time = 17.52168508
min number of rhs calls: 6
avg number of rhs calls: 358
max number of rhs calls: 5737

The speed-up is the same when considering the enthalpy system. On development, with the analytic Jacobian, we get:

Run time = 20.20451861
min number of rhs calls: 6
avg number of rhs calls: 450
max number of rhs calls: 12023

so this change is definitely a win.

[zingale]

I think that this is ready to be reviewed. Here's an updated derivation of the system used for energy.

simplified_sdc_derivations.pdf

[zingale]

Castro detonation tests, running for 250 steps.

• numerical Jacobian:

  Run time without initialization = 97.3357413

• analytic Jacobian with Microphysics on development:

  Run time without initialization = 2439.936833

• analytic Jacobian with Microphysics on sdc_jacobian before my most recent changes
  (9d097e745ab5c00620546f7e0d154df34b1a98f3):

  dies with burn failures right out of the gate in lots of different zones

• analytic Jacobian with Microphysics on sdc_jacobian after my
  changes:

  Run time without initialization = 66.74610559