post_process gas internal energy diagnostic drops pi_inf and qv, and hard-codes fluid 2

[sbryngelson]

s_write_energy_data_file (src/post_process/m_data_output.fpp:1238) accumulates the gas internal
energy as

Egint = Egint + q_prim_vf(eqn_idx%E + 2)%sf(i, j, k)*(gammas(2)*pres)*dV

The shared operator for exactly this quantity is

f_phase_internal_energy(pres, alpha, alpha_rho, gamma, pi_inf, qv) = alpha*(gamma*pres + pi_inf) + alpha_rho*qv

so the diagnostic drops both the alpha*pi_inf term and the alpha_rho*qv term, and hard-codes
fluid 2 rather than deriving which phase it means.

Consequences:

• Correct only when fluid 2 has pi_inf == 0 and qv == 0. For a stiffened-gas or an energetic
  phase in slot 2, the reported gas internal energy is wrong by alpha_2*pi_inf_2 + alpha_rho_2*qv_2.
• Silently wrong if a case orders its fluids differently, since nothing checks that fluid 2 is the
  gas. The variable name Egint is the only thing asserting it.

This is written output rather than solver state, so it does not affect the evolution - the same class
as #1773.

Suggested fix: call f_phase_internal_energy with fluid 2's coefficients, which restores the two
missing terms and makes the phase choice explicit at the call site. If fluid 2 really is a required
convention here, it deserves a comment saying so, or a validator rule.

Found while sweeping for the last equation-of-state expressions still written by hand for #1762 /
#1638. Two others remain, and both look deliberate rather than defective: the Tait isentrope applied
inline at m_riemann_solver_hllc.fpp:406-408, and the mixture-coefficient time derivative at
m_cbc.fpp:848-850.

[sbryngelson]

Narrowing this after reading the whole routine — I overstated the fluid-2 part.

s_write_energy_data_file is inherently a two-phase liquid/gas diagnostic: Vl accumulates
adv(1), Vb accumulates adv(2), and Elk/Egk split the kinetic energy on the same convention.
So gammas(2) for the gas internal energy is consistent with the rest of the routine, not a stray
hard-coded index. Withdrawing that half of the report.

The defect that stands is the dropped terms. Against

f_phase_internal_energy(pres, alpha, alpha_rho, gamma, pi_inf, qv) = alpha*(gamma*pres + pi_inf) + alpha_rho*qv

the diagnostic omits alpha_2*pi_inf_2 and alpha_rho_2*qv_2. For the gas phase pi_inf is
normally zero, so this is usually harmless in practice — but it is not guaranteed, and a stiffened or
energetic phase in slot 2 makes the reported gas internal energy wrong with no indication.

Also worth recording: sim_data appears in no test case and no example, so this routine is entirely
unexercised in-tree. Any fix has to be validated by running it directly.