fix: preserve non-ASCII tool result text

[cyphercodes]

Description

@tool results returned as dictionaries or lists currently escape CJK and emoji characters because the standard JSON serializer defaults to ASCII escaping, while Pydantic model returns keep UTF-8 text. This keeps structured tool-result text human-readable regardless of whether a tool returns a dict, list, or model.

Related Issues

Resolves #2636

Documentation PR

No documentation changes are needed for this bug fix.

Type of Change

Bug fix

Testing

• uv run --extra dev pytest tests/strands/tools/test_decorator.py -vv

• uv run --extra dev ruff format --check src/strands/tools/decorator.py tests/strands/tools/test_decorator.py

• uv run --extra dev ruff check src/strands/tools/decorator.py tests/strands/tools/test_decorator.py

• uv run --extra dev mypy src/strands/tools/decorator.py

• I ran hatch run prepare

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• Documentation changes are not needed for this bug fix
• No new docs example is needed
• My changes generate no new warnings
• Any dependent changes have been merged and published, or are not needed

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

[yatszhash]

Thanks for fixing the decorator path here! 🙏

Heads-up on scope: the provider-side json.dumps escaping that #2636 also listed — tool-result {"json": ...} content blocks and tool-call argument serialization across Anthropic / OpenAI-family providers (hit in practice by e.g. multi-agent / agent-as-tool structured output) — isn't covered by this PR. I've split it into #2660 and opened PR #2661 for it.

This PR's Resolves #2636 is fine to keep — #2660 tracks the provider-side remainder independently.