Issues: Fix JSON Serialization and add Corresponding Test

[nambili-samuel]

I enjoy contributing to projects, whether by fixing small bugs or tackling larger issues.
This small fix addresses a bug in the MCP tool (autogen_ext/tools/mcp/_base.py) where JSON serialization did not set ensure_ascii=False.

Problem:
Non-ASCII characters, including Japanese, were escaped as Unicode codepoints (e.g., "\u65e5\u672c").
Escaped characters reduce readability and can degrade LLM understanding and generation quality in non-English environments.

Solution:
Updated JSON serialization in _base.py to include ensure_ascii=False.
Added a new test (test_mcp_json_serialization.py) to verify that non-ASCII characters (e.g., Japanese text) are serialized correctly without escaping.

Files Changed:
(1) python/packages/autogen-ext/src/autogen_ext/tools/mcp/_base.py
(2) python/packages/autogen-ext/tests/tools/test_mcp_json_serialization.py

Impact:
Fixes JSON serialization for non-ASCII characters.
Ensures improved LLM compatibility in multilingual environments.

Testing:
Added unit tests for Japanese characters serialization. Verified existing functionality is unaffected.