Update to xmsconan 2.23.0 and add a Python 3.14 build

[wdolinar]

Two commits: a generator bump that regenerates CI, and the Python 3.14 matrix leg.

Update to xmsconan 2.23.0 (66a033b)

Bumps the generator from 2.21.0 and regenerates. The CI workflow picks up several template changes:

• Third-party actions pinned to commit SHAs instead of tags. A tag is a movable ref in someone else's repository, and this workflow hands its environment AQUAPI_PASSWORD_SECRET and AQUAVEO_GITHUB_TOKEN.
• flake8-tidy-imports installed in the lint job. It registers banned-modules, which the generated .flake8 sets. flake8 ignores config options no installed plugin claims, so the osgeo.* ban has been silently enforcing nothing while the job reported the same green as a run that had checked it.
• setup-python v2 → v5; xmsconan pin ==2.16.0 → >=2.23.0.
• Wheels built only in the Release leg; the Debug leg was producing wheels nothing consumed.
• The Linux job takes its Python version from the matrix rather than hardcoding it in the container image tag.

.gitignore gains .claude/worktrees/ and .worktreeinclude, plus CMakePresets.json — that file is generated, and xmscore, xmsgrid, xmsinterp and xmsextractor already ignore it, so this repo was the outlier.

Add a Python 3.14 build (d442252)

Two coupled changes; the 3.14 leg does not work without both.

Dependency bump to the versions that publish py3.14 binaries:

	before	after
xmscore	7.0.11	7.0.13
xmsgrid	9.0.10	9.1.1
xmsinterp	7.0.9	7.0.12
xmsextractor	10.0.7	10.0.10

The previous pins publish only 3.10 and 3.13, so a 3.14 leg would resolve no binary and fail at conan install. Upstream rolled 3.14 out in one wave on the newest release of each — the intermediate versions (xmscore 7.0.12, xmsgrid 9.0.11 and 9.1.0, xmsinterp 7.0.11, xmsextractor 10.0.9) do not carry it. These are the minimum versions that work, not simply the latest.

Matrix change. [ci].python_versions gains "3.14", and mac_python_versions / linux_python_versions are now named explicitly. python_versions fans out Windows only; macOS and Linux default to the single highest entry, so adding 3.14 on its own would have moved both platforms off 3.13 rather than adding a leg to either. Result: mac 3.13 + 3.14, linux 3.13 + 3.14, Windows 3.10 + 3.13 + 3.14.

Action required before merge

Fanning out mac and linux makes the generator append -py<version> to job and artifact names:

• linux job: GCC-13 (Release, Linux) → GCC-13 (Release, 3.13, Linux)
• wheel artifacts: wheel-macOS → wheel-macOS-py3.13

If branch protection on master lists the old linux job names as required status checks, they will never report and PRs will hang. The same applies to anything fetching those artifacts by exact name. I did not touch branch-protection settings.

Verification

Run on macOS arm64 (apple-clang 17):

• Release matrix builds and ctest passes 1/1 against the bumped dependencies, so the xmsgrid minor bump does not break XmGridTrace.
• A forced 3.14 pybind build produces xmsgridtrace-...-cp314-cp314-macosx_15_0_arm64.whl and passes all 19 Python tests.
• flake8 clean with the full CI plugin set, including the now-active banned-modules rule.
• xmsconan_gen + xmsconan_ci re-run with zero drift.
• ghcr.io/aquaveo/conan-gcc13-py3.14:latest confirmed to exist — the linux 3.14 leg cannot start without it.