Drop Python 3.9 support due to critical security issue with a dependency

[shachafl]

We have kept support for Python 3.9 beyond its recommended life span although it was dropped by the scientific community in April 2024 (https://numpy.org/neps/nep-0029-deprecation_policy.html).
A critical vulnerability in Dask <= 2024.8.2 forces us to drop Python 3.9 as Dask dropped support for Python 3.9 in version 2024.8.1.
After this change we will still have at least 3 working versions of Python (3.10-12).

Subtasks:

1. Remove python 3.9 in automatic GitHub Actions tests (.github/workflows/starfish-prod-ci.yml)
2. Update 4 requirements files
3. Bump python version to 3.10 in docker/Dockerfile
4. Update docs (README.rst & docs/source/installation/index.rst)
5. Update comments in examples/quick_start/plot_quick_start.py
6. Remove Python 3.9 in packaging files (pyproject.toml)
7. Make a new release by bumping minor (update CHANGELOG.md)

[Copilot]

Pull Request Overview

This PR drops support for Python 3.9 due to a critical security issue with the Dask dependency and updates various configurations and documentation to reflect the new minimum Python version.

• Updated Python version requirements in pyproject.toml to 3.10–3.12.
• Modified GitHub Actions workflows to remove Python 3.9, add macOS support, and adjust test matrices.
• Updated the quick start example, documentation, and changelog to reflect the change.

Reviewed Changes

Copilot reviewed 4 out of 11 changed files in this pull request and generated no comments.

File	Description
pyproject.toml	Updated Python version requirements and classifiers to 3.10–3.12.
examples/quick_start/plot_quick_start.py	Adjusted the documentation text to recommend Python 3.10+ (3.12 recommended).
CHANGELOG.md	Added entries for dropping Python 3.9 and other related version updates.
.github/workflows/starfish-prod-ci.yml	Replaced Python 3.9 with Python 3.10 across setup steps and expanded OS support.

Files not reviewed (7)

• README.rst: Language not supported
• docker/Dockerfile: Language not supported
• docs/source/installation/index.rst: Language not supported
• requirements/REQUIREMENTS-CI.txt: Language not supported
• requirements/REQUIREMENTS-JUPYTER.txt: Language not supported
• requirements/REQUIREMENTS-NAPARI-CI.txt: Language not supported
• starfish/REQUIREMENTS-STRICT.txt: Language not supported

[shachafl]

The security issue with Dask has been removed in the Github Security section, so no need to drop Python 3.9 yet.

Regarding the test failure on MacOS with new Apple silicon (M1+), it is probably due to slight differences in floating-point arithmetic between different hardware architectures. This could be mitigated by adding a small tolerance when comparing image registration results:

# Adding a tolerance
tolerance = 1e-4

assert np.allclose(
    expected_registered_values,
    registered_image.xarray[2, 2, 0, 40:50, 40:50],
    atol=tolerance
)

Another possible mitigation is changing float32 to float64 as mentioned in: https://stackoverflow.com/questions/71441137/np-float32-floating-point-differences-between-intel-macbook-and-m1
But I haven't tried it, and in any case, it will burden memory usage so might be less desirable.
Also, could be related to numpy update from 1.22.4 to 1.24.4 (less likely) or to 1.26.4 (not sure if tested on M1 the last update).