Varying atm option fixes rho at 1d-10

I wouldn't call this a bug, but at the Massive star Hackathon @rhdtownsend identified that we are using a fixed density floor for the outer_BC in the atmosphere integration routine when `atm_T_tau_opacity = 'varying'`:

https://github.com/MESAHub/mesa/blob/18474579ec583eb9783b09733507c8f8195c10e7/atm/private/atm_t_tau_varying.f90#L277

It's not clear why this value was chosen, but reducing it to a lower density might be warranted, and even necessary for some models where the surface extends beyond 1d-10 g/cm^2 in density. It remains to be seen.

Start with atm/private/atm_t_tau_varying.f90 at the linked outer_BC code around line 277, and inspect how the fixed 1d-10 density floor is used when atm_T_tau_opacity is 'varying'. Check whether relevant models reach densities below that floor, then determine and validate an appropriate lower value. Done means the atmosphere integration handles those models without an unjustified density cutoff.

Written by the indexing model from the issue text.