There is a potential bug in the heursitic_atmosphere function when using the full 4c radiance model.

Currently, the transm variable in the function is initialized only with the downwards components:

transm = instrument.sample(
    x_instrument, RT.wl, rhi["transm_down_dir"] + rhi["transm_down_dif"]
) 

Regardless if this is the behavior we want, it is inconsistent with the 1c implementation where the populated transm LUT variable contains the full coupled transmittance spectrum. To make this consistent:

transm = instrument.sample(
    x_instrument,
    RT.wl,
    (rhi["transm_down_dir"] + rhi["transm_down_dif"]) * (rhi["transm_up_dir"] + rhi["transm_up_dif"])
)

First pass tests show that this fixes the initial issue of poor water vapor retrievals with the 4c model: