fix(tools): stringify Enum values in legacy schema declarations

[chelsealong]

Link to Issue or Description of Change

Link to an existing issue:

• Closes: IntEnum tool parameters produce an invalid function declaration: integer values in a STRING enum #6978

Problem:

When a tool function has a parameter annotated with an enum.IntEnum (or
any Enum whose member .values are not strings), the legacy
_parse_schema_from_parameter builds a declaration with type: STRING but
copies the raw e.value into schema.enum, e.g. [1, 2]. That is invalid
by types.Schema's own contract — Schema.enum is list[str] — so:

• Building the same schema through the public constructor
  (types.Schema(type='STRING', enum=[1, 2])) raises a ValidationError.
• Serializing the ADK-built schema emits
  PydanticSerializationUnexpectedValue warnings, because the assignment
  happens via schema.enum = [...] after construction, bypassing
  validation.

Solution:

Stringify enum values (str(e.value)) when building the STRING schema,
and stringify the default the same way before comparing/assigning it. This
keeps the declaration internally consistent with the STRING type it
already declares. str-valued enums are unaffected, since str(value) == value for them (confirmed by the existing test_enums test, which still
passes unchanged).

The call-path question the issue raised (how a string-vs-name enum value
maps back onto the Python Enum when the model calls the tool) is a
separate, already-preexisting gap — there is no enum re-hydration in
FunctionTool._preprocess_args today for str enums either — and is out
of scope for this fix, which only makes the generated declaration valid.

Testing Plan

Unit Tests:

• I have added or updated unit tests for my change.
• All unit tests pass locally.

Added TestBuildFunctionDeclarationLegacy.test_int_enum in
tests/unittests/tools/test_build_function_declaration.py, which builds a
declaration for an IntEnum-typed parameter and asserts:

• schema.type == 'STRING'
• schema.enum == ['1', '2'] (strings, not ints)
• schema.default == '1'
• the resulting (type, enum) pair round-trips through types.Schema(...)
  without raising.

Verified the test fails without the fix:

$ git stash push -- src/google/adk/tools/_function_parameter_parse_util.py
$ python -m pytest tests/unittests/tools/test_build_function_declaration.py -k test_int_enum -q
...
E     assert level_schema.enum == ['1', '2']
E     AssertionError: assert [1, 2] == ['1', '2']
1 failed, 57 deselected in 0.72s
$ git stash pop

And passes with the fix:

$ python -m pytest tests/unittests/tools/test_build_function_declaration.py -k "test_int_enum or test_enums or test_enum_parameter" -q
3 passed, 55 deselected in 1.33s

Full tests/unittests/tools/ directory (2119 tests):

$ python -m pytest tests/unittests/tools/ -q
2119 passed, 1 skipped, 762 warnings in 46.53s

Manual End-to-End (E2E) Tests:

Reproduced the exact repro from the issue (from_function_with_options on
a function with an IntEnum-typed parameter, promoting the pydantic
serializer UserWarning to an error). Before the fix:

UserWarning: Pydantic serializer warnings:
  PydanticSerializationUnexpectedValue(Expected `str` - serialized value may not be as expected [field_name='enum', input_value=1, input_type=int])
  PydanticSerializationUnexpectedValue(Expected `str` - serialized value may not be as expected [field_name='enum', input_value=2, input_type=int])

After the fix, decl.parameters.properties['level'].model_dump(...) returns
{'enum': ['1', '2'], 'type': 'STRING'} with no warnings.

Checklist

• I have read the CONTRIBUTING.md document.
• I have performed a self-review of my own code.
• I have added tests that prove my fix is effective or that my feature works.
• New and existing unit tests pass locally with my changes.
• I have manually tested my changes end-to-end.
• Any dependent changes have been merged and published in downstream modules.

Additional context

Reported with AI assistance (this PR was authored by an AI coding agent);
the fix was implemented, tested, and reviewed following the reproduction
steps and root cause in the issue report.