Enforce SPEC 0

[mphoward]

We started this discussion here:

#267

but never followed up on it. I am now fully supportive of strictly adopting SPEC 0:

https://scientific-python.org/specs/spec-0000/

Maintaining backwards compatibility with Python is less of an issue given how many different package managers are available for installing specific versions. It is also a lot of work to test so many combinations and make sure they work.

SPEC 0 would mean:

• python >= 3.12
• freud >= 3.1
• gsd >= 3.3
• networkx >= 3.4
• numpy >= 2
• scipy >= 1.14
• HOOMD >= 4.8

LAMMPS is trickier because of how they name their packages. The LAMMPS conda-forge recipe also has not been updated in a long time, so I am not sure how we will test modern versions. Hence, we may leave LAMMPS pinned at the versions we currently support.

[mphoward]

@clpetix do you agree with this change? We can time enforcing it with the 0.3.0 release for your work.

[clpetix]

I am in favor of this. It is becoming unruly to maintain longterm support given how fast some of our dependencies are changing.

We should figure out how to implement the table discussed in the original discussion into our documentation to make it clear what is supported or not.