Rework the error handling exercise to be based on the expression evaluator exercise

[randomPoison]

This PR follows up from #2517, taking the error handling that was removed in that PR and reintroducing it as the error handling exercise. The core goal of the exercise is the same as before: Take some code that panics and make it return a Result, but I think this change is an improvement in a few ways:

• The new exercise is simpler, focusing on a single function with a single error case.
• It still requires that students both create an error and propagate it with ?.
• It removes the need for thiserror as an external crate, simplifying the solution and allowing it to be tested with mdbook test.
• It demonstrates that an error type can be as simple as an empty struct.
• Reusing the solution from a previous exercise should reduce the upfront cognitive burden for students, hopefully meaning they can focus more on solving the solution instead of trying to wrap their head around a parser implementation.

[djmitche]

This looks great, maybe just update the comments as noted.

[djmitche]

src/pattern-matching/exercise.rs does not contain this conditional (it just relies on / to panic).

I think that's OK, but maybe claiming it's "the original" in the comment above is inaccurate enough to bother folks (like me)

[randomPoison]

Yeah true it's not exactly the same as the solution to the earlier exercise, I deliberately added in the panic to show students where exactly the error case is. Personally I don't think that'd be too confusing for students, but let me add a speaker note calling that out to teachers so they can mention it if students are confused.