add capability to directly solve NSE with (rho, e, Ye) input

[zhichen3]

No description provided.

[zhichen3]

I did the testing using the nse_net_cell, where I loop over the part where we use (rho, e, Ye) input to find the NSE state and see the runtime, which I modified currently in this PR. I'll remove that once I clean up everything.

Seems like I can get a 30 ~ 40% speedup:

Runtime for mode = 1 : 0.111245203
-----------------------------------------------------
Runtime for mode = 2 : 0.064497065

[zhichen3]

I have notes for NSE stuff attached here. Calculation of the Jacobian stuff starts at page 24. There are other derivations which I wrote down following pynucastro appendix for self reference, in case others are interested.
nse_annotated.pdf

[zingale]

There's a circular dependency:

../../nse_solver/nse_eqns.H:6:10: error: circular header file dependency detected while including 'nse_solver.H', please check the include path [misc-header-include-cycle,-warnings-as-errors]
    6 | #include <nse_solver.H>
      |          ^
../../nse_solver/nse_solver.H:23:10: note: 'nse_eqns.H' included from here
   23 | #include <nse_eqns.H>
      |          ^
../../nse_solver/nse_check.H:14:10: note: 'nse_solver.H' included from here
   14 | #include <nse_solver.H>

[zhichen3]

I think its fixed now.