Migrate ceptr to Python 3.11–3.14

[terencelehmann]

Problem

Python 3.10 reaches end-of-life in October 2026 and will receive no further security fixes. Python 3.11 also receives security updates only. This PR migrates ceptr from >=3.10,<3.12 to >=3.11,<3.15, adding Python 3.14 support while dropping 3.10.

Changes

1. Dependencies — all bumped to the minimum version that ships a cp314 wheel:

Package	Before	After
numpy	^1.25.2	^2.3.0
scipy	^1.11.2	^1.16.1
matplotlib	^3.7.2	^3.10.0
pandas	>=2.0.3	>=2.3.3
pyarrow	^15.0.0	>=19.0.0
Pint	^0.23	^0.24
symengine	^0.9.2	^0.14.0
Cantera	>=3.1.0	>=3.2.0

2. CI — matrix extended from ['3.10', '3.11'] to ['3.11', '3.12', '3.13', '3.14']. Poetry version updated from 1.4.2 to 2.1.1.

3. Linting toolchain — Black, isort, flake8 and its six plugins (flake8-bugbear, pep8-naming, flake8-docstrings, flake8-use-fstring, flynt, flake8-polyfill) have been replaced with a single Ruff installation. The main driver is flake8-use-fstring, which has been unmaintained since July 2022 and is incompatible with flake8 7.x. Rather than upgrading flake8 and keeping six separate plugin version constraints in sync, Ruff re-implements all of them natively. It also replaces Black (via ruff format) and isort (via the I rule set), cutting the dev dependency count from 9 tools to 1. All existing lint rules are preserved with equivalent Ruff selectors; the .flake8 config file has been removed and replaced by [tool.ruff] sections in pyproject.toml. A one-time reformatting pass was applied to the codebase via commit 4fa0443.

4. New test — test_polimi2020 added, exercising a mechanism with PLOG reactions at 1 atm (plog_pressure=101325.0).

Test results

All 5 pytest tests pass on Python 3.11, 3.12, 3.13, and 3.14.

C++ output was diffed against a Python 3.11 / symengine 0.9.2 baseline for 7 mechanisms (air, LiDryer, dodecane_lu, grimech30, NUIGalway, POLIMI2020, C3MechLite). Six are byte-for-byte identical.

NUIGalway differs in 4 transport coefficient values by 1 ULP in the 8th decimal place (e.g. -1.79310920E+01 → -1.79310921E+01). This is caused by symengine 0.14 producing a slightly different CSE expression ordering, not by Python 3.14 itself. The difference is at the limit of single-precision representable accuracy and has no physical significance, but it is a change from the previously generated files and is reported here for transparency.

---

Co-authored-by: Claude Sonnet 4.6 claude-sonnet-4-6@anthropic.com