Numpy restricted to <2

[westjl]

Numpy is now up to version 2.4. Is there a reason that the numpy version is required to be numpy<2 in setup.py?

[Cameron-Van-Eck]

I spent a little time on this last week, and didn't manage to solve it. I'll put some notes here, for anyone else who picks this up in the future (possibly my future self).

All the RM synthesis tools seems to be compatible with Numpy 2+, modulo some very minor fixes (which are now on the dev branch) for deprecated features/behaviour. The problem is with the QU-fitting tool, specifically with the Bilby requirement. Upgrading to the most recent numpy release causes Bilby to crash, due to something internal to the package (not anything we're doing with Bilby). I don't really understand why or how this happens, and a quick test with a older version of Bilby didn't help.

Maybe there's some major combination of Bilby version and newer Numpy that works, or maybe we have to raise an issue with the Bilby devs, but I don't have time now for a deeper investigation. So for now, people who only want to use RM synthesis and use new Numpy releases, but QU-fitting will require more work. I'm not sure when I'll have time to return to this.