Validate dataset mixture fractions before loading

[lindicaphxag-tech]

What does this PR do?

Dataset mixtures with invalid fraction values currently fail after loading datasets, and an all-zero mixture raises an unhelpful ZeroDivisionError. This validates fractions before any dataset loading, rejects negative or non-finite values, and reports a clear error for a non-positive total.

This is an internal input-validation fix; no additional dependencies are required.

Before submitting

• This PR fixes a typo or improves the docs (not applicable).
• Did you read the contributor guideline, Pull Request section?
• Was this discussed/approved via a GitHub issue? No issue link is available for this edge case.
• Did you make sure to update the documentation with your changes? No documentation changes are needed for this internal validation fix.
• Did you write any new necessary tests?

AI writing disclosure

• No AI usage: the PR was written entirely by a human.
• AI-assisted: some parts were suggested or improved by AI, but the PR was written and reviewed by a human.
• AI-generated: the PR was mostly or fully generated by an AI tool.

Testing

• python -m pytest tests/test_cli_utils.py::TestGetDataset::test_dataset_fraction_zero_sum_raises_before_loading tests/test_cli_utils.py::TestGetDataset::test_dataset_fraction_invalid_value_raises_before_loading tests/test_cli_utils.py::TestGetDataset::test_dataset_fraction -q (5 passed)
• python -m pytest tests/test_cli_utils.py::TestGetDataset -q (15 passed; 2 pre-existing Windows NamedTemporaryFile permission failures)

Who can review?

Anyone in the community is free to review the PR once the tests have passed.

---

Note

Low Risk
Input validation only in dataset mixture loading; behavior for valid configs is unchanged aside from failing earlier on invalid fractions.

Overview
get_dataset now validates mixture fraction values before any load_dataset calls, so bad configs fail fast with clear errors instead of after expensive I/O (or with an unhelpful ZeroDivisionError when all fractions are zero).

When any dataset specifies a fraction, the check enforces: all-or-none fraction usage (unchanged), finite non-negative values, a strictly positive sum, and the existing no fractions with streaming rule. Mixture sizing logic after load is unchanged.

Tests assert load_dataset is never called for all-zero sums and for negative, NaN, or infinite fractions.

^{Reviewed by Cursor Bugbot for commit 4384e32. Bugbot is set up for automated code reviews on this repo. Configure here.}