Making ast.walk 220x Faster

Reflex's AI code generator encountered performance issues with its custom Python linter, specifically with the `ast.walk` function which was taking ~285 nanoseconds per node. By systematically removing generators and optimizing Python-level code, the team achieved a 2x improvement, but ultimately rewrote the function in Rust using PyO3 bindings to achieve a 220x speedup in overall performance.