ast: testing and fuzzing

The parser needs (thorough) testing, currently there is none. Ideas include trying to match the generated AST data to a known format (yaml, etc) and/or rendering the generated AST data back to text (already supported by render.zig) and comparing the result to the original.

It would also be nice to fuzz the parser. This will take some more thought.

Start by examining the parser and the generated AST, then read render.zig to understand the existing AST-to-text path. Define a focused round-trip or known-format test strategy before tackling fuzzing, since the issue leaves the scope open. Done means the parser has thorough automated coverage and a considered fuzzing approach.

Written by the indexing model from the issue text.