Add [[ablation]] sweep syntax for eval TOML configs

[snimu]

Expand cartesian products of parameter values into eval configs automatically. Supports top-level sweep keys and env_args sweep keys, fixed env_args alongside swept ones (with overlap guard), multiple independent ablation blocks, and coexistence with [[eval]].

Description

Type of Change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Test improvement

Testing

• All existing tests pass when running uv run pytest locally.
• New tests have been added to cover the changes

Checklist

• My code follows the style guidelines of this project as outlined in AGENTS.md
• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

Note

Medium Risk
Touches core config-loading logic used by prime eval, so schema/validation mistakes could break existing TOML workflows, though changes are well-scoped and covered by new tests.

Overview
Adds support for [[ablation]] blocks in eval TOML configs, expanding a cartesian product of [ablation.sweep] (and [ablation.sweep.env_args]) values into multiple concrete eval configs alongside existing [[eval]] entries.

Updates load_toml_config to parse/validate ablation sections, apply global defaults to generated configs, guard against invalid fields and overlapping fixed vs swept env_args, and continue resolving endpoints_path relative to the config. Documentation and the eval skill guide now describe the new syntax, and the test suite adds comprehensive coverage for expansion behavior and error cases.

^{Written by Cursor Bugbot for commit 4d8dd11. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}