fix: unused step and step_end results

[makcandrov]

This PR fixes a bug introduced by #759:

• The result of step is read but not written to the interpreter result. If the function returns something other than Continue, it will lead to an infinite loop in the interpreter.
• The result of step_end is not being read at all, which means the execution cannot be halted by the inspector from this function.

[rakita]

Hey, I wanted to remove the return type from step and step_end as you can set that value directly inside Inspector interpreter.instruction_result = result.

Would be interested to get your feedback on that approach?

[makcandrov]

Yes, I thought about it too - it definitely makes sense.

However, the current implementation is a confusing mix between the two and silently breaks some inspectors. That's why I suggested reverting to the original behavior.

If you want to take a look, I tried implementing the new behavior here (I wanted to open the PR in my fork it ended up in the main repo - sorry)

[rakita]

Yes, I thought about it too - it definitely makes sense.

However, the current implementation is a confusing mix between the two and silently breaks some inspectors. That's why I suggested reverting to the original behavior.

If you want to take a look, I tried implementing the new behavior here (I wanted to open the PR in my fork it ended up in the main repo - sorry)

Yeah, the current main approach is definitely a footgun.

I like this other PR more: #804 as it does not reuse the InstructionResult as returns (that felt bad in past, it should have been enum of some kind) and it is more straightforward on what changes inside aka interpreter.instruction_result.

[makcandrov]

Closing in favor of #804