Particles splits script

[robjmcgibbon]

This PR updates a couple of things relating to the particle splitting script

• Look for .txt files instead of .hdf5, and don't complain if the file is empty
• The original script was written with numpy 1.xx, but numpy 2 was giving an overflow error when trying to add a np.int64 to an array of python ints
• SWIFT outputs negative numbers for progenitor trees. I'm not convinced we're handling this properly

[robjmcgibbon]

Particles with 64 splits can have a SplitTrees value which is negative. This is because SWIFT outputs the SplitTrees values as a signed long. This was causing issues since python has unbounded integers, so I now convert all negative integers to their equivalent unsigned version.

Since SWIFT no longer output negative progenitor IDs (which was a bug), I have removed the code we put in place to handle such particles. This is just reverting #59

I don't think any more changes are required. I'm currently testing this on the L0200N3008 box.

• Finish testing

[robjmcgibbon]

I looked at a few particles with > 64 splits in the L0200N3008 and they appear to have been processed correctly. For reference I was looking at snapshots 53/54, as that's where the first overflow appears. The region is [150.87459943 93.73696643 98.33375143]

We need to rerun the particle splitting (and HBT itself) for snapshot 38 onwards