Use previous ion populations as the first guess for the NLTE solver

[chvogl]

Is your feature request related to a problem? Please describe. Currently, the first guess for the ion populations in the NLTE solver is that everything is singly ionised. This is a sensible guess for the first iteration. However, in subsequent iterations it is much better to use the ion populations from the previous iteration.

Steps:

• Define previous iteration properties for ion number densities and add them to the properties collection
• Add them as inputs to the NLTESolver
• make sure that the store method is called (https://github.com/tardis-sn/tardis/blob/f7dd18ab5c100fdee9d6e7c22ba2e62852386a68/tardis/simulation/base.py#L357C34-L357C34)