fix: handle Optional[str] in pre_parse_json to prevent incorrect type coercion

[Maanik23]

Fixes #381

Summary

The pre_parse_json method in func_metadata.py uses field_info.annotation is not str to decide whether to attempt JSON parsing of string values. This check fails for union types like Optional[str] (str | None) because the annotation object is str | None, not str.

This causes string values like "1.2" or '{"key": "value"}' to be parsed via json.loads into float/dict, which then fails Pydantic validation with Input should be a valid string.

Reproducer (from issue)

@server.tool()
async def getinfo(cid: Optional[str] = Field(..., description="xxx.")) -> bool:
    ...

# Client calls with: {"cid": "1.2"}
# Result: ValidationError - Input should be a valid string [input_value=1.2, input_type=float]

Changes

src/mcp/server/mcpserver/utilities/func_metadata.py

• Added _annotation_accepts_str() helper that correctly identifies str-only unions (Optional[str], str | None) as string-accepting annotations
• Uses is_union_origin (already imported) and get_origin/get_args (already imported) for introspection
• Preserves existing behavior for unions with other complex types (e.g. str | list[str] still gets JSON pre-parsing)

tests/server/mcpserver/test_func_metadata.py

• Added test_optional_str_preserves_json_string: numeric strings, JSON objects, and JSON arrays all stay as str when annotation is Optional[str]
• Added test_pipe_none_str_preserves_json_string: same coverage for str | None syntax
• Added test_optional_str_runtime_validation: end-to-end call_fn_with_arg_validation with Optional[str] param

Backward compatibility

• No change for str annotations (already handled by PR Avoiding to parse string arguments containing valid JSON if the argument annotation is str #1113)
• No change for str | list[str] unions (still parses JSON arrays into lists)
• Fixes Optional[str] and str | None to correctly preserve string values
• All 36 existing tests in test_func_metadata.py continue to pass