fix(extract-rust): mark #[test]/#[tokio::test] fns is_test so they are filtered

[moofone]

Closes #855.

Summary

• Inline Rust test functions — #[test], #[tokio::test], #[async_std::test], #[actix_rt::test], #[test_case::case] — inside a regular .rs file are now marked is_test = true on the Function node, so the store.c is_test != 1 filter excludes them from graph/agent context. Previously they leaked in because test detection was file-path-based only (cbm_is_test_file: *_test.rs / test_*).
• Adds rust_def_is_test(decorators) (mirrors rust_cfg_qualified_name's decorator scan), set on def in extract_func_def's Rust branch. extract_decorators stores attribute_item text in its bracketed form ("#[test]", "#[tokio::test]", "#[tokio::test(...)]"), so the matcher is substring-based and robust to argument lists.
• Rust-scoped per the issue. Free functions are covered, including #[cfg(test)] mod tests { #[tokio::test] async fn … }. impl/trait methods can follow in a separate change.

Verification

• scripts/test.sh → 5831 passed (clean ASan + UBSan build). New regression extract_rust_test_attr_marks_is_test_issue855 covers #[test], #[tokio::test], and a negative (non-test) case.
• Red → green: with detection disabled, the new test fails exactly at ASSERT(sync == 1 && "#[test] fn is_test") and is the only failure (5830 passed, 1 failed); restored → green.
• scripts/lint.sh (clang-tidy / cppcheck / clang-format) is not available in my local environment, so CI is the authority for it; the new code mirrors the surrounding rust_cfg_qualified_name style.

Notes

• DCO-signed. Happy to adjust the attribute set or the extraction-hook point per maintainer feedback on fix(extract-rust): inline test fns (#[test]/#[tokio::test]) are not marked is_test — leak into graph/agent context #855.

[DeusData]

Thank you — this is a correct, well-scoped fix: detecting #[test]/#[tokio::test]/#[async_std::test]/etc. attributes and marking those functions is_test, which flows end-to-end into the node properties and the store queries that exclude tests from graph/agent context. The decorator substring-matching mirrors the existing rust_cfg pattern, and your test (on a non-test path so is_test can only come from the attribute detection) is a genuine non-vacuous guard. Merged as bf3657f; closes #855. Deferring impl/trait test methods to a follow-up was the right call. Appreciated!