Possible difference in behaviour with CPU and GPU for the event timestep_ended

Sharing the following observation from Nicolas

> it seems that we have a different behaviour with CPU and GPU solvers when it comes to getting the data from PyFluent.... I'm using the callback timestep_ended and when I get the data from there (at the end of each timestep), the data only changes every 50 timesteps or so. Same code on CPU solver works fine, we get different data at each timestep.

Start with the PyFluent timestep_ended callback and reproduce the observation with both CPU and GPU solvers. Compare the data returned at each timestep, especially whether GPU results change only every roughly 50 timesteps; the issue is done when the difference is understood and the expected per-timestep behavior is verified or documented.

Written by the indexing model from the issue text.