[Proposal] ABNF Fuzzing

[gluax]

💥 Proposal

With the upcoming parser re-write, we no longer have the pest file to make the language grammar easily determinable. However, since we still have a written ABNF notation I think an ideal way to test human-readable grammar is through fuzzing. ABNF notation is machine-readable, so we can have a text file with our ABNF rules that are read, and then generate valid Leo programs from those rules. This could make sure we conform to the grammar we have written. As a bonus, there may already be libraries that implement ABNF fuzzing that we could just leverage. The end goal of this would be to have a CI that generates test cases based on the ABNF notation to ensure we meet our expected grammar.