[data] fix MiniMax tool_extractor to return content when no tool call parses

[he-yufeng]

What does this PR do?

MiniMaxM1ToolUtils.tool_extractor and MiniMaxM2ToolUtils.tool_extractor return an empty list when the tool-call wrapper tag is present but no valid call parses out of it. M1 skips lines that fail json.loads (via continue), and M2 finds no <invoke> block, so results ends up empty. An empty list reads downstream as "a tool call with zero functions" rather than "plain content", so the original assistant text is dropped.

This returns the original content in that case, matching SeedToolUtils (fixed in #10408), Qwen35ToolUtils and LFM2ToolUtils, which already use return results if results else content. DefaultToolUtils, QwenToolUtils and Gemma4ToolUtils already handle it (they return content on a parse failure, or append one call per regex match), so only the two MiniMax extractors needed the change.

Added tests/data/test_tool_utils.py covering both the empty case (tag present, nothing parses, returns content) and the valid case (a real call still returns the FunctionCall list).

Before submitting

• Did you read the contributor guideline?
• Did you write any new necessary tests?

[gemini-code-assist]

Code Review

This pull request updates the tool extractors for MiniMaxM1 and MiniMaxM2 in src/llamafactory/data/tool_utils.py to return the original content instead of an empty list when no valid tool calls are parsed. Additionally, corresponding unit tests have been added in tests/data/test_tool_utils.py to verify both successful parsing and fallback behavior. No review comments were provided, and the changes look correct and well-tested.

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.

[he-yufeng]

Closing this in favor of #10599, which is the same fix (return the raw content when the wrapper tag is present but nothing parses inside it) on a fresh branch and is the one currently under review. No point keeping two identical PRs open. Thanks!