Add delay to activation of IFP tallies.

[GuySten]

Description

Currently, IFP tallies are scored in active batches but because they contain data from ifp_n_generations before they contain data from inactive batches (where the fission source distribution might not be converged).
Instead of making the user remember to increase the number of inactive batches when using IFP tallies this PR propose the elegant solution of activating the IFP tallies later (by ifp_n_generation).

This PR also further simplify the IFP implementation.

Checklist

• I have performed a self-review of my own code
• I have run clang-format (version 15) on any C++ source files (if applicable)
• I have followed the style guidelines for Python source files (if applicable)
• I have made corresponding changes to the documentation (if applicable)
• I have added tests that prove my fix is effective or that my feature works (if applicable)

[JoffreyDorville]

Hi @GuySten, thanks for the effort! I will probably need some time to get back to it more in depth, but I really like what you did in terms of simplification and cleaning. Do you see any impact in terms of runtime performance?

[GuySten]

The motivation for the simplifications is just cleaner code.
I did not check performance difference.

[GuySten]

This PR is now ready for review.

[nuclearkevin]

Is there a specific reason as to why the IFP tests required a regold? Just asking as we'll need to regold our IFP tests in Cardinal the next time we do an OpenMC submodule update (after this is merged), and will need a reason for SQA purposes.

[GuySten]

Because this change decrease the statistics of the IFP tallies.
They are now activated with a delay of ifp_n_generation so they will not rely on data from inactive cycles (which might be unsafe).

Another reason is that in one test I increased the number of particles to compensate the decrease of statistics.