Generalize with a `ProvableProgram` trait

In the `aptos-lc` crate, the `merkle` and `ratchet` modules have similar functions called `generate_stdin` and `generate_keys`.

Would it be better if we abstracted those in a `ProvableProgram` trait? It's not so clear because they don't share much code and it would risk bloating the code with boilerplate

Start in the aptos-lc crate by reading the merkle and ratchet modules and comparing their generate_stdin and generate_keys functions. Determine whether their shared behavior justifies a ProvableProgram trait without adding boilerplate; done means the abstraction question has a clearly documented decision and any agreed refactor is scoped.

Written by the indexing model from the issue text.