Add node-level blocker support to ancestor propagation, refactor logic, add tests, bump version

[jeromychu23]

Motivation

• Add support for an optional node-level block_key that stops ancestor-based target propagation when a blocking event is found with a later target timestamp.
• Refactor the propagation implementation to separate pure row processing from the Python-facing wrapper for easier testing and clarity.
• Bump package and recipe versions to publish the new behavior and update the recipe build metadata.

Description

• Introduce has_blocker_on_node_after_target to detect blocker events on nodes and wire it into traversal termination logic.
• Split propagate_target_from_ancestor into an internal propagate_target_rows function that operates on &[PyDict] and a thin #[pyfunction] wrapper that accepts a DataFrame and optional block_key parameter.
• Rename and simplify helper functions (matches_target_key -> matches_kv) and adjust calls to composite_key and choose_candidate_by_plan to accept consistent argument types.
• Add multiple unit tests under #[cfg(test)] covering blocker termination, self-node blockers, and parent-node blockers.
• Bump crate version in Cargo.toml to 0.1.3 and update the conda recipe recipe/meta.yaml version to 0.1.3 while adjusting the build number and test commands to match the new parameter names and signature.

Testing

• Ran cargo test which executed the new propagate_target_rows unit tests and they passed.
• Updated recipe test commands in recipe/meta.yaml to call the extended propagate_target_from_ancestor signature; these are included as automated package tests in the recipe.

---

Codex Task