Convergence check for t_r is not working

[kslong]

In trying to look a bit more at questions associated with convergence, I find that t_r_old and t_r are always identical and hence all cells appear to be converged by that measure, even though t_r is clearly changing.

This may be a reappearance of #110

[kslong]

Unlike #110 this problem does not go away in single processor mode. Looking back over regression tests for the last year, this problem has been their all along. Shame on us for not seeing it earlier.

The problem here is that t_r_old and t_r are identical when the check is made.

[kslong]

As an explanation here. The reason one cannot treat t_r_old the same way one does t_e_old, is that t_e is updated in ionization c after all of the photons have been translated for the current cycle. So in ionization.c one can do the update. But t_r_old needs to be done immediately before t_r is calculated from the frequency distribution.