Support Python 3.12-3.14 in Conda builds

[harryswift01]

Summary

This PR makes the Conda package actually installable across all supported Python versions (3.12, 3.13, 3.14), previously the recipe only ever built and published a single Python 3.14 variant regardless of what meta.yaml claimed to support.

Changes

Declare the Python build matrix in the Conda recipe:

• Added conda-recipe/conda_build_config.yaml declaring python: ["3.12", "3.13", "3.14"], the standard conda-forge/bioconda way of expressing a multi-version build matrix
• Changed meta.yaml's host/test sections to a bare, unconstrained python so conda-build's variant system actually substitutes the target version, an explicit range there (the previous python >=3.12) silently blocks this substitution
• Aligned run: python >=3.12,<3.15 with pyproject.toml's requires-python

Publish all Python/platform combinations on release:

• Simplified release.yaml's conda job back to an os-only matrix, since a single conda build now produces all three Python variants per platform automatically
• Updated the build/upload steps to handle multiple package outputs per job instead of assuming one

Simplify weekly regression testing to match:

• Dropped the now-redundant python-version matrix from weekly-regression.yaml, each OS job now builds and tests all three variants in one run
• Bumped timeout-minutes from 30 to 90 to account for the extra build/test work per job

Fix stale documentation:

• Updated installation.rst/development.rst from "Python 3.12 or later" to "Python 3.12, 3.13, or 3.14" to match the actual supported range

Impact

• Releases now publish 9 packages (3 platforms × 3 Python versions) instead of 3, fixing installs for anyone not on Python 3.14
• Verified end-to-end with real local builds (not just rendering): all three Python variants build and pass all 18 regression tests, both via conda_build_config.yaml's default matrix and via explicit --python overrides
• No behavioural change to SIST itself, this is entirely packaging/CI

[jimboid]

Fixes a major annoyance with code not installing on different versions of python despite them bring listed in meta.yaml/pyptoject.toml in the conda release chain. Happy to merge, you will want to release at some point to reflect the change.