Add structured output format validation helpers

[ECNUACRush]

Summary

Fixes #265.

This PR adds evaluator wrappers for structured-output validation:

• require_json_output
• require_xml_output
• require_regex_match
• require_format

When validation fails, the wrapper returns score=0.0 and includes the parse or validation error in side_info, so GEPA's reflection step can learn from format failures.

This is related to #379/#384, but intentionally limited to evaluator-side format failure feedback. It does not implement rejected-proposal memory.

Tests

• python -m pytest -q
• python -m ruff check src/gepa/utils/format_validation.py src/gepa/utils/__init__.py tests/test_format_validation.py docs/docs/guides/faq.md
• python -m pyright --pythonversion 3.12 src/gepa/utils/format_validation.py
• mkdocs build -f docs/mkdocs.yml

[semanticdiff-com]

Review changes with  

Changed Files

File	Status
docs/docs/guides/faq.md	Unsupported file format
src/gepa/utils/__init__.py	0% smaller
src/gepa/utils/format_validation.py	0% smaller
tests/test_format_validation.py	0% smaller

[mvanhorn]

Heads up: I opened #387 for this same issue a few minutes after you filed this, without seeing yours first. Yours is more complete (XML support plus the FAQ docs), so I'm closing mine in favor of it.

One small edge case my tests covered that might be worth folding in: a require_format validator that raises an exception rather than returning False. My version catches that and treats it as a format failure (score 0 with the error in feedback) instead of letting it propagate out of the evaluator. Take it or leave it. Thanks for picking this up.

[ECNUACRush]

Thanks for the heads up, and no worries. I appreciate you closing yours. @mvanhorn

Good catch on the require_format exception case. I’ll add handling for it.