dispersion free distribution

Following discussions with @gnigrell and looking at the numbers with @seyue we find the distribution covariance generated with dispersion_free is different from the one with zero longitudinal emittance.

Just pasting the code in chatGPT it proposes an explanation as to the reason, could someone check it and possibly if @gnigrell you have the rights update the master while @kskoufar is on holyday?

Thanks Yann