Add support for numpy 2.x

[aalexandersson]

Is your proposal related to a problem?

pseudopeople has the install requirement "numpy<2.0.0", in setup.py#L39 which does not allow numpy>=2.0.0.

Describe the solution you'd like

I suggest changing setup.py#L39 to "numpy>=2.0.0",

Describe alternatives you've considered

An alternative is to keep the current version for those users who do not need the numpy upgrade. Another alternative is to add branch logic based on the version of Python, so that only pseudopeople in Python 3.11 and later will use "numpy>=2.0.0". For comparison, Splink in pyproject.toml uses this branching logic:

python = ">=3.8.0,<4.0.0"
jsonschema = ">=3.2"
# 1.3.5 is the last version supporting py 3.7.1
pandas = ">1.3.5"
duckdb = ">=0.9.2"
sqlglot = ">=13.0.0"
altair = "^5.0.1"
Jinja2 = ">=3.0.3"

# need to manually specify numpy versions suitable for CI
# 1.24.4 works with python 3.8, but not 3.12
numpy = [
    # version is minimum valid with above listed pandas version
    {version=">=1.17.3", python = "<3.12"},
    {version=">=1.26.0", python = ">=3.12"},
]

Additional context

Splink 4.x and Pandas 2.2.2+ support numpy 2.0. pseudopeople should stop referring to Python 3.8 because Python 3.8 is unsupported and therefore a security risk, see issue 469.

[aflaxman]

Thanks for flagging this @aalexandersson. We are enhancing our approach to large-scale testing right now, and using the enhanced testing approach to see if there are any pitfalls with Python 3.12 and numpy 2.x will be a great way to see if it works.

[aalexandersson]

Thank you. For your testing, you might find helpful the implementation in Splink v4.0.6. See pull #2518.

[aflaxman]

Completed by PR #553