improve  Doppler averaging:  right now  it is  done in  XMDS with  box-like
integration. Small reprogramming could use different weights for points see
my own  (i.e Eugeniy's)  lectures, and  make it trapezoid  or even  Si with
great improvement on precision.

move maxwell probabilities away from inner  loops. It need to be calculated
only once.

connect cell/atom temperature with density of atoms in vapor cell