test(cel): pin stdlib-vs-custom and namespace-vs-scalar collision pre…

[aviavissar]

Closes #3

Pins CEL PDP collision-precedence that already exists in code. Tests only; no behavior change.

• Custom size vs stdlib: matching stdlib overload runs (size('hello') == 5); custom runs only when stdlib has no matching form (size(42) == 777).
• Bag tree: namespace wins over scalar on the same path, in both insert orders.

Supersedes #58 (same commit; branch was renamed).

[praxis-bot]

praxis-bot review

The Rust changes (activation.rs, resolver.rs) are well-structured: the collision-precedence tests pin real behavior with clear assertion messages, and the insert_key utility correctly mirrors production's split logic. The doc comment additions on with_functions document important dispatch semantics.

One concern with scope and one with test coverage below.

Findings: 1 Large, 1 Medium

[praxis-bot]

[Large] This 906-line Hebrew explainer document is unrelated to the stated PR purpose ("pin CEL PDP collision-precedence... Tests only; no behavior change"). It adds a standalone HTML page with external font dependencies, inline CSS, and relative cross-repo links (../../praxis/filter/...) that will break if either repo's directory structure changes.

Split this into its own PR so the test-pinning change can be reviewed and merged independently. A documentation PR also benefits from its own review pass for link accuracy and content correctness.

[praxis-bot]

[Medium] The new doc comment states "If two custom setups register the same name, the later one wins" but the test added below (custom_size_stdlib_overload_then_int_fallback) only covers stdlib-vs-custom dispatch. Add a test that registers two custom functions with the same name and signature, then asserts the later registration's body runs. That pins the claim made here and prevents a silent regression if the CEL crate changes registration semantics.