revisit the transverse update

[zingale]

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.

[zingale]

I think that this would actually match what we do with MHD

[zingale]

this is implemented in #1920

[zingale]

note that the bug fix on the reset species with density was done separately now in #2120