Bump max Python version to 3.13

[maread99]

Bumps max Python version to 3.13.

Updates:

• workflows to use 3.13 (including upper version of test matrix).
• pyproject.toml project classifiers (as declared to PyPI).

[maread99]

Tests failing because numpy 1.26 won't build with python 3.13...

The minimum version of pandas that we support is currently 1.5.
pandas 1.5 requires numpy < 2.0
It seems that numpy 1.26.4 (the last version of v1 numpy) does not support Python 3.13

So, seems that to support Python 3.13 we'll have to drop support for pandas v1 and numpy v1... (?)

[maread99]

.... all tests pass if we drop the 'pandas >= 1.5' restriction.

@gerrymanoim, would you be happy to lose the minimum pandas 1.5 restraint in order that we can support Python 3.13? (Obviously users could still use Python 3.10 through 3.12 with pandas 1.5 and numpy 1.26).

Cheers
Marcus

[gerrymanoim]

Could we change

        python-version: ["3.10", "3.12"]
        include:
          - python-version: "3.10"
            requirements_file: requirements_dev.txt
          - python-version: "3.12"
            requirements_file: requirements_minpandas.txt

to be

        python-version: ["3.10", "3.12"]
        include:
          - python-version: "3.10"
            requirements_file: requirements_minpandas.txt
          - python-version: "3.12"
            requirements_file: requirements_dev.txt

instead? That way we're still testing older pands/python but not blocking anyone on py13?

I generally prefer to follow to the NEP support/drop schedule https://numpy.org/neps/nep-0029-deprecation_policy.html#drop-schedule, but I'm open to changing this if its painful.

[maread99]

We could do, although if we're going to explicitly support 3.13 then should we lose the pandas>=1.5 declaration in pyproject.toml and change it to simply pandas? (Otherwise there would be a conflict in what we're saying is supported.)

Shall I swap the requirements files over, as you suggest, and change the dependency in pyproject.toml to simply pandas. That way the requirements would be congruous with the supported Python versions and we'd still be implicitly ensuring support for pandas / numpy v1 with at least the earliest supported Python version.

[gerrymanoim]

Yep - I think that makes sense to me - thanks!