CI updates for caching

[j-adamczyk]

Changes

CI updates, particularly for more efficient caching. Also covers #498 by fixing dependencies bounds.

Checklist before requesting a review

• Docstrings added/updated in public functions and classes
• Tests added, reasonable test coverage (at least ~90%, make test-coverage)
• Sphinx docs added/updated and render properly (make docs and see docs/_build/index.html)

[Copilot]

Pull request overview

This PR updates CI infrastructure for more efficient caching and addresses dependency bound issues (issue #498). The changes include upgrading package versions, removing unused dependencies, updating NumPy API calls for compatibility, and improving the GitHub Actions caching strategy.

Key Changes

• Updated dependency version constraints to be more specific (numpy >=1.26.4, pandas >=1.5.3, scikit-learn >=1.6.0, scipy >=1.12.0, rdkit >=2023.9.6)
• Removed numba and llvmlite dependencies (verified unused in codebase)
• Updated test code to use current NumPy API (np.issubdtype instead of deprecated np.isdtype, np.bool_ instead of np.bool)
• Improved CI caching to use dataset file hashes as cache keys instead of generic "datasets" key
• Added backward compatibility for root_mean_squared_error in older scikit-learn versions

Reviewed changes

Copilot reviewed 18 out of 20 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
pyproject.toml	Updated dependency bounds with more specific minimum versions; removed setuptools and numba dependencies
uv.lock	Updated package versions including numpy 2.3.5→2.4.0, librt 0.7.4→0.7.5, mistune 3.1.4→3.2.0, nbclient 0.10.3→0.10.4, psutil 7.1.3→7.2.0, pyparsing 3.2.5→3.3.1; removed numba and llvmlite
tests/fingerprints/getaway.py	Limited test dataset size to 30 items to reduce computation time
tests/distances/common.py	Made scipy sparse class assertion more flexible to handle version differences
tests/applicability_domain/*.py	Updated NumPy API calls from np.isdtype(x, np.bool) to np.issubdtype(x, np.bool_) across multiple test files
tests/applicability_domain/distance_to_centroid.py	Fixed test function names from test_knn_* to test_distance_to_centroid_*
skfp/metrics/multioutput.py	Added try/except import for root_mean_squared_error with fallback for older scikit-learn versions
skfp/distances/tanimoto.py	Removed obsolete Numba-related comment
ruff.toml	Added target-version configuration and reordered ignore list
benchmarking/benchmark.py	Added strict=False to zip() calls for Python 3.10+ compatibility
Makefile	Minor whitespace cleanup
.github/workflows/tests.yml	Improved caching strategy using dataset file hashes as cache key and simplified venv setup