Correct equation linearization issues

[jorekart]

Correct equation linearization issues

Depends on #41 (automatic equation verificator). Please merge that one first.

This PR fixes the linearization problems in models/model600/mod_elt_matrix_fft.f90
found by the equation verificator introduced in #41. That tool linearizes the
model-600 weak forms symbolically and compares the result, term by term, with the
hand-written residual and Jacobian. The finding numbers below refer to its
util/equation_codegen/JOREK_FINDINGS.md. There is one commit per finding.

RHS fixes (these change the converged solution)

• Finding 10: rhs_ij_k(var_rhoimp): the toroidal channel was missing the
  impurity shock-capturing diffusivity D_par_imp_sc_num*tau_sc.
• Finding 16: rhs_ij(var_T): the parallel pressure convection used the
  neutral density rn0 where it should use rimp0.

Amat fixes

• Finding 3: the single-temperature diamagnetic-viscosity tangent was a
  factor of two too large.
• Finding 5: the rho equation's neutral sources were missing the
  impurity/alpha_e(T) contributions.
• Finding 6: the vpar time term is now linearized consistently
  (B²·dvpar/dt + ½·vpar·dB²/dt).
• Finding 7: the tgnum_vpar terms were missing the BB2/psi
  derivatives.
• Finding 8: the toroidal channel of the visco_par_par tangent was
  missing.
• Finding 11: the tgnum_T/Ti/Te tangents had the wrong power of BigR.
• Finding 12: three smaller omissions in the energy tangents: the vpar
  kinetic-coupling term, dZK*_prof_drho, and the recombination sink.
• Finding 13: the ionization-energy tangents were missing the theta
  factor.
• Finding 14: some terms used alpha_e/alpha_imp where the _bis form
  is needed.
• Finding 15: the T/Te tangents were missing several ionization, radiation
  and friction-source derivatives, and one term used the uncorrected
  densities.
• Finding 17: amat_n(var_T,var_vpar) was missing the impurity pressure.

Known remaining differences

• Findings 4 and 9 (inward pinch): the pinch is not differentiated with
  respect to psi, and the vpar pinch tangent has the wrong sign. These are
  left to the pinch developer. @szmate00 , could you look a this?
• Findings 1 and 2 (impurity pressure in the tauIC diamagnetic tangents):
  these are accepted as known differences, because impurities and
  tauIC /= 0 cannot be run together. That combination should be rejected at
  input time; it is not enforced yet.