Being able to specify recursion limits in stack depth, rather than in bytes, would improve diagnostics and reliability of exceeding the limit.

We could also use it to quickly limit execution in fuzzing tests