fix: accept markdown-fenced JSON in validate_json_format

[Chessing234]

Summary

• Strip json / fences before json.loads, matching IFEvalG.

Test plan

• Unit tests for bare and fenced JSON
• GPU_TESTS=bypass

[gemini-code-assist]

Code Review

This pull request updates the validate_json_format function in open_instruct/if_functions.py to strip markdown code fences before parsing JSON, and adds comprehensive unit tests in open_instruct/test_if_functions.py. The reviewer feedback suggests replacing the chained .removeprefix() and .removesuffix() calls with a robust, case-insensitive regular expression to better handle mixed-case language identifiers (e.g., jSoN) and avoid potential parsing failures.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Using chained .removeprefix() and .removesuffix() calls can be fragile and lead to bugs. For example, if the model outputs a mixed-case identifier like jSoN, the specific prefixes (json, Json, JSON) won't match, but the fallback .removeprefix() for ```` will match and strip only the backticks, leaving the malformed string jSoN\n{"a": 1} which fails JSON parsing.\n\nUsing a single case-insensitive regular expression is much more robust, cleaner, and handles any whitespace or casing variations automatically.

    match = re.match(r"^```(?:json)?\s*(.*?)\s*```$", text.strip(), re.DOTALL | re.IGNORECASE)
    value = match.group(1) if match else text.strip()