code_review: tolerate double-encoded enum tool args

[suhaibmujahid]

Models occasionally send enum tool arguments double-encoded, e.g. the literal '"exclude"' (quotes included) for the search_text/search_identifier 'tests' parameter, which failed pydantic Literal validation and crashed the tool call with a ValidationError.

Add a reusable bugbug.tools.core.validators module exposing strip_enum_quotes and a StripEnumQuotes BeforeValidator, used as Annotated[Literal[...], StripEnumQuotes] so any LLM-fed enum param can opt in. The 'tests' and 'langs' params now strip surrounding quotes/whitespace before the Literal check. The JSON schema sent to Anthropic is unchanged (still emits the enum), so strict-mode validation is preserved.

Fixes #6140

[Copilot]

Pull request overview

This PR makes the code-review toolchain more tolerant of a common LLM failure mode where enum-valued tool arguments are double-encoded (e.g. the literal '"exclude"'), by stripping surrounding quotes/whitespace before Literal[...] validation. This prevents pydantic ValidationErrors while keeping the generated JSON schema enums intact for strict-mode validation.

Changes:

• Added a reusable bugbug.tools.core.validators module with strip_enum_quotes and a StripEnumQuotes BeforeValidator.
• Updated code-review Searchfox tool argument types (langs, tests) to opt into quote-stripping via Annotated[..., StripEnumQuotes].
• Added unit/regression tests covering both the validator behavior and the specific search_text / search_identifier regression from #6140.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 1 comment.

File	Description
bugbug/tools/core/validators.py	Introduces a reusable BeforeValidator to unwrap double-encoded enum-like strings.
bugbug/tools/code_review/langchain_tools.py	Applies the validator to Searchfox tool enum params (langs, tests) via Annotated.
tests/test_tools_core_validators.py	Adds focused unit tests for quote-stripping and schema preservation.
tests/test_code_review.py	Adds async regression tests ensuring tool invocation accepts double-encoded args and passes normalized values to the client.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.