revisit the transverse update

Right now, we do the transverse correction to the conserved state + p. This way we don't need to use the EOS to get p on the interface when we go back to the primitive state for the Riemann solve.

But the EOS is pretty cheap now (with the templating), that it might be cleaner and more robust to update the conserved system, derive (rho e) after the correction and handle it in the same dual energy fashion as we do elsewhere, and then use ctoprim to get the new corrected primitive edge state.

Start by tracing the transverse correction to the conserved state and p, then compare it with the dual-energy handling used elsewhere and the ctoprim path for the corrected primitive edge state. Done means the correction uses the revised conserved-state and dual-energy approach while producing the corrected primitive edge state for the Riemann solve.