fix: retain model names containing interval-like text in evaluation

[student-shivamm]

Fixes #311

When models is omitted, evaluate treats model-hi-2025-v2 as an interval column and silently omits its score. Explicit model selection returns the expected score. Anchor interval matching to the numeric suffix so interval-like text inside model names is retained, while integer, fractional and scientific-notation interval columns remain excluded.

Regression tests compare inferred and explicit selection in pandas and Polars and assert known MAE values. The same 24-case matrix has 16 failures and 8 passing controls on upstream, and passes with the patch.

Validation: evaluation/loss suites plus regression module: 190 passed, one skipped (upstream Windows distributed-test restriction). Separate data/validation/processing plus regression run: 399 passed, ten skipped. All configured pinned pre-commit hooks pass on the changed files. Direct no-incremental mypy with runtime dependencies reports the same 73 existing errors on upstream and the candidate; the isolated configured hook passes. No cross-platform or distributed CI result is claimed.

AI assistance was used for investigation, implementation and validation. The reproduction and recorded local results were checked; acceptance remains the maintainers' decision.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you sign our Contributor License Agreement before we can accept your contribution.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}