Add git-based dynamic versioning with setuptools-scm

[EiffL]

Summary

• Replace hardcoded version = "0.0.1rc1" with setuptools-scm, which derives versions automatically from git tags
• Expose jax_galsim.__version__ at runtime via a generated _version.py
• Update CI to fetch full git history (fetch-depth: 0) so setuptools-scm can resolve the version, and use submodules: recursive in the checkout step instead of a manual git submodule update

How it works

setuptools-scm reads git tags at build/install time:

• At a tag like v0.0.1rc1 → version is 0.0.1rc1
• N commits past a tag → version is 0.0.1rc1.devN+g<hash>
• No tag found → falls back to 0.0.1rc1.dev0 (configured in pyproject.toml)

Action needed after merge

To anchor the versioning, a maintainer should create an initial SemVer tag on the merge commit:

git tag v0.0.1rc1
git push origin v0.0.1rc1

We recommend SemVer (MAJOR.MINOR.PATCH) over CalVer for this project because:

• As a library, downstream code depends on the API — SemVer communicates compatibility (breaking vs non-breaking changes)
• Scientific computing users need clear signals about whether an upgrade is safe
• The 0.x.y convention explicitly signals "pre-stable, API may change"

Also, I don't think setuptools-scm works with CalVer

[beckermr]

This version string

0.0.1rc1.dev0

confuses me since "rc" means release candidate and the patch version is for patches.

Maybe we should adopt

0.1.0.dev0

as the fallback?

[beckermr]

Also, I don't think setuptools-scm works with CalVer

What do you mean by this?

[beckermr]

See calver-by-date here: https://setuptools-scm.readthedocs.io/en/latest/extending/#available-implementations

[codspeed-hq]

Merging this PR will degrade performance by 30.63%

❌ 1 regressed benchmark
✅ 30 untouched benchmarks

⚠️ Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Benchmark	BASE	HEAD	Efficiency
❌	test_benchmark_spergel_calcfluxrad[run]	1 ms	1.5 ms	-30.63%

---

_{Comparing setup-dynamic-versioning (d41c90d) with main (04d22c6)}

[EiffL]

Ok excellent, happy to use CalVer then. Not sure if we should tag the branch first, for after merging the code to main.

[beckermr]

No need to tag anything before or after merging. The existing tag for the oja paper may need adjusting. We can handle that later.

Once we switch the fallback version to 0.1.0.dev0, I think this is good to go.

[beckermr]

If we switch to calver we may want to change the default version. We can do that in a future PR.